asterisk/res/res_stasis.c

1068 lines
29 KiB
C
Raw Normal View History

/*
* Asterisk -- An open source telephony toolkit.
*
* Copyright (C) 2012 - 2013, Digium, Inc.
*
* David M. Lee, II <dlee@digium.com>
*
* See http://www.asterisk.org for more information about
* the Asterisk project. Please do not directly contact
* any of the maintainers of this project for assistance;
* the project provides a web site, mailing lists and IRC
* channels for your use.
*
* This program is free software, distributed under the terms of
* the GNU General Public License Version 2. See the LICENSE file
* at the top of the source tree.
*/
/*! \file
*
* \brief Stasis application support.
*
* \author David M. Lee, II <dlee@digium.com>
*
* <code>res_stasis.so</code> brings together the various components of the
* Stasis application infrastructure.
*
* First, there's the Stasis application handler, stasis_app_exec(). This is
* called by <code>app_stasis.so</code> to give control of a channel to the
* Stasis application code from the dialplan.
*
* While a channel is in stasis_app_exec(), it has a \ref stasis_app_control
* object, which may be used to control the channel.
*
* To control the channel, commands may be sent to channel using
* stasis_app_send_command() and stasis_app_send_async_command().
*
* Alongside this, applications may be registered/unregistered using
* stasis_app_register()/stasis_app_unregister(). While a channel is in Stasis,
* events received on the channel's topic are converted to JSON and forwarded to
* the \ref stasis_app_cb. The application may also subscribe to the channel to
* continue to receive messages even after the channel has left Stasis, but it
* will not be able to control it.
*
* Given all the stuff that comes together in this module, it's been broken up
* into several pieces that are in <code>res/stasis/</code> and compiled into
* <code>res_stasis.so</code>.
*/
/*** MODULEINFO
<support_level>core</support_level>
***/
#include "asterisk.h"
ASTERISK_FILE_VERSION(__FILE__, "$Revision$")
#include "asterisk/astobj2.h"
#include "asterisk/callerid.h"
#include "asterisk/module.h"
#include "asterisk/stasis_app_impl.h"
#include "asterisk/stasis_channels.h"
#include "asterisk/stasis_bridges.h"
#include "asterisk/stasis_message_router.h"
#include "asterisk/strings.h"
#include "stasis/app.h"
#include "stasis/control.h"
/*! Time to wait for a frame in the application */
#define MAX_WAIT_MS 200
/*!
* \brief Number of buckets for the Stasis application hash table. Remember to
* keep it a prime number!
*/
#define APPS_NUM_BUCKETS 127
/*!
* \brief Number of buckets for the Stasis application hash table. Remember to
* keep it a prime number!
*/
#define CONTROLS_NUM_BUCKETS 127
/*!
* \brief Stasis application container.
*/
struct ao2_container *apps_registry;
struct ao2_container *app_controls;
struct ao2_container *app_bridges;
/*! \brief Message router for the channel caching topic */
struct stasis_message_router *channel_router;
/*! \brief Message router for the bridge caching topic */
struct stasis_message_router *bridge_router;
/*! AO2 hash function for \ref app */
static int app_hash(const void *obj, const int flags)
{
const struct app *app = obj;
const char *name = flags & OBJ_KEY ? obj : app_name(app);
return ast_str_hash(name);
}
/*! AO2 comparison function for \ref app */
static int app_compare(void *lhs, void *rhs, int flags)
{
const struct app *lhs_app = lhs;
const struct app *rhs_app = rhs;
const char *lhs_name = app_name(lhs_app);
const char *rhs_name = flags & OBJ_KEY ? rhs : app_name(rhs_app);
if (strcmp(lhs_name, rhs_name) == 0) {
return CMP_MATCH | CMP_STOP;
} else {
return 0;
}
}
/*! AO2 hash function for \ref stasis_app_control */
static int control_hash(const void *obj, const int flags)
{
const struct stasis_app_control *control = obj;
const char *id = flags & OBJ_KEY ?
obj : stasis_app_control_get_channel_id(control);
return ast_str_hash(id);
}
/*! AO2 comparison function for \ref stasis_app_control */
static int control_compare(void *lhs, void *rhs, int flags)
{
const struct stasis_app_control *lhs_control = lhs;
const struct stasis_app_control *rhs_control = rhs;
const char *lhs_id = stasis_app_control_get_channel_id(lhs_control);
const char *rhs_id = flags & OBJ_KEY ?
rhs : stasis_app_control_get_channel_id(rhs_control);
if (strcmp(lhs_id, rhs_id) == 0) {
return CMP_MATCH | CMP_STOP;
} else {
return 0;
}
}
struct stasis_app_control *stasis_app_control_create(struct ast_channel *chan)
{
return control_create(chan);
}
struct stasis_app_control *stasis_app_control_find_by_channel(
const struct ast_channel *chan)
{
if (chan == NULL) {
return NULL;
}
return stasis_app_control_find_by_channel_id(
ast_channel_uniqueid(chan));
}
struct stasis_app_control *stasis_app_control_find_by_channel_id(
const char *channel_id)
{
return ao2_find(app_controls, channel_id, OBJ_KEY);
}
/*! AO2 hash function for bridges container */
static int bridges_hash(const void *obj, const int flags)
{
const struct ast_bridge *bridge = obj;
const char *id = flags & OBJ_KEY ?
obj : bridge->uniqueid;
return ast_str_hash(id);
}
/*! AO2 comparison function for bridges container */
static int bridges_compare(void *lhs, void *rhs, int flags)
{
const struct ast_bridge *lhs_bridge = lhs;
const struct ast_bridge *rhs_bridge = rhs;
const char *lhs_id = lhs_bridge->uniqueid;
const char *rhs_id = flags & OBJ_KEY ?
rhs : rhs_bridge->uniqueid;
if (strcmp(lhs_id, rhs_id) == 0) {
return CMP_MATCH | CMP_STOP;
} else {
return 0;
}
}
struct ast_bridge *stasis_app_bridge_find_by_id(
const char *bridge_id)
{
return ao2_find(app_bridges, bridge_id, OBJ_KEY);
}
/*! \brief Typedef for blob handler callbacks */
typedef struct ast_json *(*channel_blob_handler_cb)(struct ast_channel_blob *);
/*! \brief Callback to check whether an app is watching a given channel */
static int app_watching_channel_cb(void *obj, void *arg, int flags)
{
struct app *app = obj;
char *uniqueid = arg;
return app_is_watching_channel(app, uniqueid) ? CMP_MATCH : 0;
}
/*! \brief Get a container full of apps that are interested in the specified channel */
static struct ao2_container *get_apps_watching_channel(const char *uniqueid)
{
struct ao2_container *watching_apps;
char *uniqueid_dup;
RAII_VAR(struct ao2_iterator *,watching_apps_iter, NULL, ao2_iterator_destroy);
ast_assert(uniqueid != NULL);
uniqueid_dup = ast_strdupa(uniqueid);
watching_apps_iter = ao2_callback(apps_registry, OBJ_MULTIPLE, app_watching_channel_cb, uniqueid_dup);
watching_apps = watching_apps_iter->c;
if (!ao2_container_count(watching_apps)) {
return NULL;
}
ao2_ref(watching_apps, +1);
return watching_apps_iter->c;
}
/*! \brief Typedef for callbacks that get called on channel snapshot updates */
typedef struct ast_json *(*channel_snapshot_monitor)(
struct ast_channel_snapshot *old_snapshot,
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
struct ast_channel_snapshot *new_snapshot,
const struct timeval *tv);
static struct ast_json *simple_channel_event(
const char *type,
struct ast_channel_snapshot *snapshot,
const struct timeval *tv)
{
return ast_json_pack("{s: s, s: o, s: o}",
"type", type,
"timestamp", ast_json_timeval(*tv, NULL),
"channel", ast_channel_snapshot_to_json(snapshot));
}
static struct ast_json *channel_created_event(
struct ast_channel_snapshot *snapshot,
const struct timeval *tv)
{
return simple_channel_event("ChannelCreated", snapshot, tv);
}
static struct ast_json *channel_destroyed_event(
struct ast_channel_snapshot *snapshot,
const struct timeval *tv)
{
return ast_json_pack("{s: s, s: o, s: i, s: s, s: o}",
"type", "ChannelDestroyed",
"timestamp", ast_json_timeval(*tv, NULL),
"cause", snapshot->hangupcause,
"cause_txt", ast_cause2str(snapshot->hangupcause),
"channel", ast_channel_snapshot_to_json(snapshot));
}
static struct ast_json *channel_state_change_event(
struct ast_channel_snapshot *snapshot,
const struct timeval *tv)
{
return simple_channel_event("ChannelStateChange", snapshot, tv);
}
/*! \brief Handle channel state changes */
static struct ast_json *channel_state(
struct ast_channel_snapshot *old_snapshot,
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
struct ast_channel_snapshot *new_snapshot,
const struct timeval *tv)
{
struct ast_channel_snapshot *snapshot = new_snapshot ? new_snapshot : old_snapshot;
if (!old_snapshot) {
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
return channel_created_event(snapshot, tv);
} else if (!new_snapshot) {
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
return channel_destroyed_event(snapshot, tv);
} else if (old_snapshot->state != new_snapshot->state) {
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
return channel_state_change_event(snapshot, tv);
}
return NULL;
}
static struct ast_json *channel_dialplan(
struct ast_channel_snapshot *old_snapshot,
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
struct ast_channel_snapshot *new_snapshot,
const struct timeval *tv)
{
RAII_VAR(struct ast_json *, json, NULL, ast_json_unref);
/* No Newexten event on cache clear */
if (!new_snapshot) {
return NULL;
}
/* Empty application is not valid for a Newexten event */
if (ast_strlen_zero(new_snapshot->appl)) {
return NULL;
}
if (old_snapshot && ast_channel_snapshot_cep_equal(old_snapshot, new_snapshot)) {
return NULL;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
return ast_json_pack("{s: s, s: o, s: s, s: s, s: o}",
"type", "ChannelDialplan",
"timestamp", ast_json_timeval(*tv, NULL),
"dialplan_app", new_snapshot->appl,
"dialplan_app_data", new_snapshot->data,
"channel", ast_channel_snapshot_to_json(new_snapshot));
}
static struct ast_json *channel_callerid(
struct ast_channel_snapshot *old_snapshot,
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
struct ast_channel_snapshot *new_snapshot,
const struct timeval *tv)
{
RAII_VAR(struct ast_json *, json, NULL, ast_json_unref);
/* No NewCallerid event on cache clear or first event */
if (!old_snapshot || !new_snapshot) {
return NULL;
}
if (ast_channel_snapshot_caller_id_equal(old_snapshot, new_snapshot)) {
return NULL;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
return ast_json_pack("{s: s, s: o, s: i, s: s, s: o}",
"type", "ChannelCallerId",
"timestamp", ast_json_timeval(*tv, NULL),
"caller_presentation", new_snapshot->caller_pres,
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
"caller_presentation_txt", ast_describe_caller_presentation(
new_snapshot->caller_pres),
"channel", ast_channel_snapshot_to_json(new_snapshot));
}
channel_snapshot_monitor channel_monitors[] = {
channel_state,
channel_dialplan,
channel_callerid
};
static int app_send_cb(void *obj, void *arg, int flags)
{
struct app *app = obj;
struct ast_json *msg = arg;
app_send(app, msg);
return 0;
}
static void sub_channel_snapshot_handler(void *data,
struct stasis_subscription *sub,
struct stasis_topic *topic,
struct stasis_message *message)
{
RAII_VAR(struct ao2_container *, watching_apps, NULL, ao2_cleanup);
struct stasis_cache_update *update = stasis_message_data(message);
struct ast_channel_snapshot *new_snapshot = stasis_message_data(update->new_snapshot);
struct ast_channel_snapshot *old_snapshot = stasis_message_data(update->old_snapshot);
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
/* Pull timestamp from the new snapshot, or from the update message
* when there isn't one. */
const struct timeval *tv = update->new_snapshot ? stasis_message_timestamp(update->new_snapshot) : stasis_message_timestamp(message);
int i;
watching_apps = get_apps_watching_channel(new_snapshot ? new_snapshot->uniqueid : old_snapshot->uniqueid);
if (!watching_apps) {
return;
}
for (i = 0; i < ARRAY_LEN(channel_monitors); ++i) {
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = channel_monitors[i](old_snapshot, new_snapshot, tv);
if (msg) {
ao2_callback(watching_apps, OBJ_NODATA, app_send_cb, msg);
}
}
}
static void distribute_message(struct ao2_container *apps, struct ast_json *msg)
{
ao2_callback(apps, OBJ_NODATA, app_send_cb, msg);
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
static void sub_channel_blob_handler(void *data,
struct stasis_subscription *sub,
struct stasis_topic *topic,
struct stasis_message *message)
{
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
RAII_VAR(struct ao2_container *, watching_apps, NULL, ao2_cleanup);
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
struct ast_channel_blob *obj = stasis_message_data(message);
if (!obj->snapshot) {
return;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = stasis_message_to_json(message);
if (!msg) {
return;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
watching_apps = get_apps_watching_channel(obj->snapshot->uniqueid);
if (!watching_apps) {
return;
}
distribute_message(watching_apps, msg);
}
/*!
* \brief In addition to running ao2_cleanup(), this function also removes the
* object from the app_controls container.
*/
static void control_unlink(struct stasis_app_control *control)
{
if (!control) {
return;
}
ao2_unlink_flags(app_controls, control,
OBJ_POINTER | OBJ_UNLINK | OBJ_NODATA);
ao2_cleanup(control);
}
struct ast_bridge *stasis_app_bridge_create(const char *type)
{
struct ast_bridge *bridge;
int capabilities, flags = AST_BRIDGE_FLAG_MERGE_INHIBIT_FROM | AST_BRIDGE_FLAG_MERGE_INHIBIT_TO
| AST_BRIDGE_FLAG_SWAP_INHIBIT_FROM | AST_BRIDGE_FLAG_SWAP_INHIBIT_TO
| AST_BRIDGE_FLAG_TRANSFER_PROHIBITED;
if (ast_strlen_zero(type) || !strcmp(type, "mixing")) {
capabilities = AST_BRIDGE_CAPABILITY_1TO1MIX |
AST_BRIDGE_CAPABILITY_MULTIMIX |
AST_BRIDGE_CAPABILITY_NATIVE;
flags |= AST_BRIDGE_FLAG_SMART;
} else if (!strcmp(type, "holding")) {
capabilities = AST_BRIDGE_CAPABILITY_HOLDING;
} else {
return NULL;
}
bridge = ast_bridge_base_new(capabilities, flags);
if (bridge) {
ao2_link(app_bridges, bridge);
}
return bridge;
}
void stasis_app_bridge_destroy(const char *bridge_id)
{
struct ast_bridge *bridge = stasis_app_bridge_find_by_id(bridge_id);
if (!bridge) {
return;
}
ao2_unlink(app_bridges, bridge);
ast_bridge_destroy(bridge);
}
int app_send_start_msg(struct app *app, struct ast_channel *chan,
int argc, char *argv[])
{
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
RAII_VAR(struct ast_channel_snapshot *, snapshot, NULL, ao2_cleanup);
struct ast_json *json_args;
int i;
ast_assert(chan != NULL);
/* Set channel info */
snapshot = ast_channel_snapshot_create(chan);
if (!snapshot) {
return -1;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = ast_json_pack("{s: s, s: [], s: o}",
"type", "StasisStart",
"args",
"channel", ast_channel_snapshot_to_json(snapshot));
if (!msg) {
return -1;
}
/* Append arguments to args array */
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
json_args = ast_json_object_get(msg, "args");
ast_assert(json_args != NULL);
for (i = 0; i < argc; ++i) {
int r = ast_json_array_append(json_args,
ast_json_string_create(argv[i]));
if (r != 0) {
ast_log(LOG_ERROR, "Error appending start message\n");
return -1;
}
}
app_send(app, msg);
return 0;
}
int app_send_end_msg(struct app *app, struct ast_channel *chan)
{
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
RAII_VAR(struct ast_channel_snapshot *, snapshot, NULL, ao2_cleanup);
ast_assert(chan != NULL);
/* Set channel info */
snapshot = ast_channel_snapshot_create(chan);
if (snapshot == NULL) {
return -1;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = ast_json_pack("{s: s, s: o}",
"type", "StasisEnd",
"channel", ast_channel_snapshot_to_json(snapshot));
if (!msg) {
return -1;
}
app_send(app, msg);
return 0;
}
void stasis_app_control_execute_until_exhausted(struct ast_channel *chan, struct stasis_app_control *control)
{
while (!control_is_done(control)) {
int command_count = control_dispatch_all(control, chan);
if (command_count == 0 || ast_channel_fdno(chan) == -1) {
break;
}
}
}
/*! /brief Stasis dialplan application callback */
int stasis_app_exec(struct ast_channel *chan, const char *app_name, int argc,
char *argv[])
{
SCOPED_MODULE_USE(ast_module_info->self);
RAII_VAR(struct app *, app, NULL, ao2_cleanup);
RAII_VAR(struct stasis_app_control *, control, NULL, control_unlink);
int res = 0;
ast_assert(chan != NULL);
app = ao2_find(apps_registry, app_name, OBJ_KEY);
if (!app) {
ast_log(LOG_ERROR,
"Stasis app '%s' not registered\n", app_name);
return -1;
}
if (!app_is_active(app)) {
ast_log(LOG_ERROR,
"Stasis app '%s' not active\n", app_name);
return -1;
}
control = control_create(chan);
if (!control) {
ast_log(LOG_ERROR, "Allocated failed\n");
return -1;
}
ao2_link(app_controls, control);
res = app_send_start_msg(app, chan, argc, argv);
if (res != 0) {
ast_log(LOG_ERROR,
"Error sending start message to '%s'\n", app_name);
return res;
}
if (app_add_channel(app, chan)) {
ast_log(LOG_ERROR, "Error adding listener for channel %s to app %s\n", ast_channel_name(chan), app_name);
return -1;
}
while (!control_is_done(control)) {
RAII_VAR(struct ast_frame *, f, NULL, ast_frame_dtor);
int r;
int command_count;
if (stasis_app_get_bridge(control)) {
/* Bridge is handling channel frames */
control_wait(control);
control_dispatch_all(control, chan);
continue;
}
r = ast_waitfor(chan, MAX_WAIT_MS);
if (r < 0) {
ast_debug(3, "%s: Poll error\n",
ast_channel_uniqueid(chan));
break;
}
command_count = control_dispatch_all(control, chan);
if (command_count > 0 && ast_channel_fdno(chan) == -1) {
/* Command drained the channel; wait for next frame */
continue;
}
if (r == 0) {
/* Timeout */
continue;
}
f = ast_read(chan);
if (!f) {
ast_debug(3,
"%s: No more frames. Must be done, I guess.\n",
ast_channel_uniqueid(chan));
break;
}
switch (f->frametype) {
case AST_FRAME_CONTROL:
if (f->subclass.integer == AST_CONTROL_HANGUP) {
/* Continue on in the dialplan */
ast_debug(3, "%s: Hangup\n",
ast_channel_uniqueid(chan));
control_continue(control);
}
break;
default:
/* Not handled; discard */
break;
}
}
app_remove_channel(app, chan);
res = app_send_end_msg(app, chan);
if (res != 0) {
ast_log(LOG_ERROR,
"Error sending end message to %s\n", app_name);
return res;
}
return res;
}
int stasis_app_send(const char *app_name, struct ast_json *message)
{
RAII_VAR(struct app *, app, NULL, ao2_cleanup);
app = ao2_find(apps_registry, app_name, OBJ_KEY);
if (!app) {
/* XXX We can do a better job handling late binding, queueing up
* the call for a few seconds to wait for the app to register.
*/
ast_log(LOG_WARNING,
"Stasis app '%s' not registered\n", app_name);
return -1;
}
app_send(app, message);
return 0;
}
static int cleanup_cb(void *obj, void *arg, int flags)
{
struct app *app = obj;
if (!app_is_finished(app)) {
return 0;
}
ast_verb(1, "Cleaning up application '%s'\n", app_name(app));
return CMP_MATCH;
}
/*!
* \brief Clean up any old apps that we don't need any more.
*/
static void cleanup(void)
{
ao2_callback(apps_registry, OBJ_MULTIPLE | OBJ_NODATA | OBJ_UNLINK,
cleanup_cb, NULL);
}
int stasis_app_register(const char *app_name, stasis_app_cb handler, void *data)
{
RAII_VAR(struct app *, app, NULL, ao2_cleanup);
SCOPED_LOCK(apps_lock, apps_registry, ao2_lock, ao2_unlock);
app = ao2_find(apps_registry, app_name, OBJ_KEY | OBJ_NOLOCK);
if (app) {
app_update(app, handler, data);
} else {
app = app_create(app_name, handler, data);
if (app) {
ao2_link_flags(apps_registry, app, OBJ_NOLOCK);
} else {
return -1;
}
}
/* We lazily clean up the apps_registry, because it's good enough to
* prevent memory leaks, and we're lazy.
*/
cleanup();
return 0;
}
void stasis_app_unregister(const char *app_name)
{
RAII_VAR(struct app *, app, NULL, ao2_cleanup);
if (!app_name) {
return;
}
app = ao2_find(apps_registry, app_name, OBJ_KEY);
if (!app) {
ast_log(LOG_ERROR,
"Stasis app '%s' not registered\n", app_name);
return;
}
app_deactivate(app);
/* There's a decent chance that app is ready for cleanup. Go ahead
* and clean up, just in case
*/
cleanup();
}
void stasis_app_ref(void)
{
ast_module_ref(ast_module_info->self);
}
void stasis_app_unref(void)
{
ast_module_unref(ast_module_info->self);
}
/*! \brief Callback to check whether an app is watching a given bridge */
static int app_watching_bridge_cb(void *obj, void *arg, int flags)
{
struct app *app = obj;
char *uniqueid = arg;
return app_is_watching_bridge(app, uniqueid) ? CMP_MATCH : 0;
}
/*! \brief Get a container full of apps that are interested in the specified bridge */
static struct ao2_container *get_apps_watching_bridge(const char *uniqueid)
{
struct ao2_container *watching_apps;
char *uniqueid_dup;
RAII_VAR(struct ao2_iterator *,watching_apps_iter, NULL, ao2_iterator_destroy);
ast_assert(uniqueid != NULL);
uniqueid_dup = ast_strdupa(uniqueid);
watching_apps_iter = ao2_callback(apps_registry, OBJ_MULTIPLE, app_watching_bridge_cb, uniqueid_dup);
watching_apps = watching_apps_iter->c;
if (!ao2_container_count(watching_apps)) {
return NULL;
}
ao2_ref(watching_apps, +1);
return watching_apps_iter->c;
}
/*! Callback used to remove an app's interest in a bridge */
static int remove_bridge_cb(void *obj, void *arg, int flags)
{
app_remove_bridge(obj, arg);
return 0;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
static struct ast_json *simple_bridge_event(
const char *type,
struct ast_bridge_snapshot *snapshot,
const struct timeval *tv)
{
return ast_json_pack("{s: s, s: o, s: o}",
"type", type,
"timestamp", ast_json_timeval(*tv, NULL),
"bridge", ast_bridge_snapshot_to_json(snapshot));
}
static struct ast_json *simple_bridge_channel_event(
const char *type,
struct ast_bridge_snapshot *bridge_snapshot,
struct ast_channel_snapshot *channel_snapshot,
const struct timeval *tv)
{
return ast_json_pack("{s: s, s: o, s: o, s: o}",
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
"type", type,
"timestamp", ast_json_timeval(*tv, NULL),
"bridge", ast_bridge_snapshot_to_json(bridge_snapshot),
"channel", ast_channel_snapshot_to_json(channel_snapshot));
}
static void sub_bridge_snapshot_handler(void *data,
struct stasis_subscription *sub,
struct stasis_topic *topic,
struct stasis_message *message)
{
RAII_VAR(struct ao2_container *, watching_apps, NULL, ao2_cleanup);
struct stasis_cache_update *update = stasis_message_data(message);
struct ast_bridge_snapshot *new_snapshot = stasis_message_data(update->new_snapshot);
struct ast_bridge_snapshot *old_snapshot = stasis_message_data(update->old_snapshot);
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
const struct timeval *tv = update->new_snapshot ? stasis_message_timestamp(update->new_snapshot) : stasis_message_timestamp(message);
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
watching_apps = get_apps_watching_bridge(new_snapshot ? new_snapshot->uniqueid : old_snapshot->uniqueid);
if (!watching_apps || !ao2_container_count(watching_apps)) {
return;
}
if (!new_snapshot) {
RAII_VAR(char *, bridge_id, ast_strdup(old_snapshot->uniqueid), ast_free);
/* The bridge has gone away. Create the message, make sure no apps are
* watching this bridge anymore, and destroy the bridge's control
* structure */
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = simple_bridge_event("BridgeDestroyed", old_snapshot, tv);
ao2_callback(watching_apps, OBJ_NODATA, remove_bridge_cb, bridge_id);
stasis_app_bridge_destroy(old_snapshot->uniqueid);
} else if (!old_snapshot) {
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = simple_bridge_event("BridgeCreated", old_snapshot, tv);
}
if (!msg) {
return;
}
distribute_message(watching_apps, msg);
}
/*! \brief Callback used to merge two containers of applications */
static int list_merge_cb(void *obj, void *arg, int flags)
{
/* remove any current entries for this app */
ao2_find(arg, obj, OBJ_POINTER | OBJ_UNLINK | OBJ_NODATA | OBJ_MULTIPLE);
/* relink as the only entry */
ao2_link(arg, obj);
return 0;
}
/*! \brief Merge container src into container dst without modifying src */
static void update_apps_list(struct ao2_container *dst, struct ao2_container *src)
{
ao2_callback(src, OBJ_NODATA, list_merge_cb, dst);
}
/*! \brief Callback for adding to an app's bridges of interest */
static int app_add_bridge_cb(void *obj, void *arg, int flags)
{
app_add_bridge(obj, arg);
return 0;
}
/*! \brief Add interest in the given bridge to all apps in the container */
static void update_bridge_interest(struct ao2_container *apps, const char *bridge_id)
{
RAII_VAR(char *, bridge_id_dup, ast_strdup(bridge_id), ast_free);
ao2_callback(apps, OBJ_NODATA, app_add_bridge_cb, bridge_id_dup);
}
static void sub_bridge_merge_handler(void *data,
struct stasis_subscription *sub,
struct stasis_topic *topic,
struct stasis_message *message)
{
RAII_VAR(struct ao2_container *, watching_apps_to, NULL, ao2_cleanup);
RAII_VAR(struct ao2_container *, watching_apps_from, NULL, ao2_cleanup);
RAII_VAR(struct ao2_container *, watching_apps_all, ao2_container_alloc(1, NULL, NULL), ao2_cleanup);
struct ast_bridge_merge_message *merge = stasis_message_data(message);
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
RAII_VAR(struct ast_json *, blob, NULL, ast_json_unref);
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
const struct timeval *tv = stasis_message_timestamp(message);
watching_apps_to = get_apps_watching_bridge(merge->to->uniqueid);
if (watching_apps_to) {
update_apps_list(watching_apps_all, watching_apps_to);
}
watching_apps_from = get_apps_watching_bridge(merge->from->uniqueid);
if (watching_apps_from) {
update_bridge_interest(watching_apps_from, merge->to->uniqueid);
update_apps_list(watching_apps_all, watching_apps_from);
}
if (!ao2_container_count(watching_apps_all)) {
return;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = ast_json_pack("{s: s, s: o, s: o, s: o}",
"type", "BridgeMerged",
"timestamp", ast_json_timeval(*tv, NULL),
"bridge", ast_bridge_snapshot_to_json(merge->to),
"bridge_from", ast_bridge_snapshot_to_json(merge->from));
if (!msg) {
return;
}
distribute_message(watching_apps_all, msg);
}
static void sub_bridge_enter_handler(void *data,
struct stasis_subscription *sub,
struct stasis_topic *topic,
struct stasis_message *message)
{
RAII_VAR(struct ao2_container *, watching_apps_channel, NULL, ao2_cleanup);
RAII_VAR(struct ao2_container *, watching_apps_bridge, NULL, ao2_cleanup);
RAII_VAR(struct ao2_container *, watching_apps_all, ao2_container_alloc(1, NULL, NULL), ao2_cleanup);
struct ast_bridge_blob *obj = stasis_message_data(message);
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
watching_apps_bridge = get_apps_watching_bridge(obj->bridge->uniqueid);
if (watching_apps_bridge) {
update_apps_list(watching_apps_all, watching_apps_bridge);
}
watching_apps_channel = get_apps_watching_channel(obj->channel->uniqueid);
if (watching_apps_channel) {
update_bridge_interest(watching_apps_channel, obj->bridge->uniqueid);
update_apps_list(watching_apps_all, watching_apps_channel);
}
if (!ao2_container_count(watching_apps_all)) {
return;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = simple_bridge_channel_event("ChannelEnteredBridge", obj->bridge,
obj->channel, stasis_message_timestamp(message));
distribute_message(watching_apps_all, msg);
}
static void sub_bridge_leave_handler(void *data,
struct stasis_subscription *sub,
struct stasis_topic *topic,
struct stasis_message *message)
{
RAII_VAR(struct ao2_container *, watching_apps_bridge, NULL, ao2_cleanup);
struct ast_bridge_blob *obj = stasis_message_data(message);
RAII_VAR(struct ast_json *, msg, NULL, ast_json_unref);
watching_apps_bridge = get_apps_watching_bridge(obj->bridge->uniqueid);
if (!watching_apps_bridge) {
return;
}
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
msg = simple_bridge_channel_event("ChannelLeftBridge", obj->bridge,
obj->channel, stasis_message_timestamp(message));
distribute_message(watching_apps_bridge, msg);
}
static int load_module(void)
{
int r = 0;
apps_registry =
ao2_container_alloc(APPS_NUM_BUCKETS, app_hash, app_compare);
if (apps_registry == NULL) {
return AST_MODULE_LOAD_FAILURE;
}
app_controls = ao2_container_alloc(CONTROLS_NUM_BUCKETS,
control_hash, control_compare);
if (app_controls == NULL) {
return AST_MODULE_LOAD_FAILURE;
}
app_bridges = ao2_container_alloc(CONTROLS_NUM_BUCKETS,
bridges_hash, bridges_compare);
if (app_bridges == NULL) {
return AST_MODULE_LOAD_FAILURE;
}
channel_router = stasis_message_router_create(ast_channel_topic_all_cached());
if (!channel_router) {
return AST_MODULE_LOAD_FAILURE;
}
r |= stasis_message_router_add(channel_router, stasis_cache_update_type(), sub_channel_snapshot_handler, NULL);
Update events to use Swagger 1.3 subtyping, and related aftermath This patch started with the simple idea of changing the /events data model to be more sane. The original model would send out events like: { "stasis_start": { "args": [], "channel": { ... } } } The event discriminator was the field name instead of being a value in the object, due to limitations in how Swagger 1.1 could model objects. While technically sufficient in communicating event information, it was really difficult to deal with in terms of client side JSON handling. This patch takes advantage of a proposed extension[1] to Swagger which allows type variance through the use of a discriminator field. This had a domino effect that made this a surprisingly large patch. [1]: https://groups.google.com/d/msg/wordnik-api/EC3rGajE0os/ey_5dBI_jWcJ In changing the models, I also had to change the swagger_model.py processor so it can handle the type discriminator and subtyping. I took that a big step forward, and using that information to generate an ari_model module, which can validate a JSON object against the Swagger model. The REST and WebSocket generators were changed to take advantage of the validators. If compiled with AST_DEVMODE enabled, JSON objects that don't match their corresponding models will not be sent out. For REST API calls, a 500 Internal Server response is sent. For WebSockets, the invalid JSON message is replaced with an error message. Since this took over about half of the job of the existing JSON generators, and the .to_json virtual function on messages took over the other half, I reluctantly removed the generators. The validators turned up all sorts of errors and inconsistencies in our data models, and the code. These were cleaned up, with checks in the code generator avoid some of the consistency problems in the future. * The model for a channel snapshot was trimmed down to match the information sent via AMI. Many of the field being sent were not useful in the general case. * The model for a bridge snapshot was updated to be more consistent with the other ARI models. Another impact of introducing subtyping was that the swagger-codegen documentation generator was insufficient (at least until it catches up with Swagger 1.2). I wanted it to be easier to generate docs for the API anyways, so I ported the wiki pages to use the Asterisk Swagger generator. In the process, I was able to clean up many of the model links, which would occasionally give inconsistent results on the wiki. I also added error responses to the wiki docs, making the wiki documentation more complete. Finally, since Stasis-HTTP will now be named Asterisk REST Interface (ARI), any new functions and files I created carry the ari_ prefix. I changed a few stasis_http references to ari where it was non-intrusive and made sense. (closes issue ASTERISK-21885) Review: https://reviewboard.asterisk.org/r/2639/ git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@393529 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-07-03 16:32:41 +00:00
/* TODO: This could be handled a lot better. Instead of subscribing to
* the one caching topic and filtering out messages by channel id, we
* should have individual caching topics per-channel, with a shared
* back-end cache. That would simplify a lot of what's going on right
* here.
*/
r |= stasis_message_router_add(channel_router, ast_channel_user_event_type(), sub_channel_blob_handler, NULL);
r |= stasis_message_router_add(channel_router, ast_channel_varset_type(), sub_channel_blob_handler, NULL);
r |= stasis_message_router_add(channel_router, ast_channel_dtmf_end_type(), sub_channel_blob_handler, NULL);
r |= stasis_message_router_add(channel_router, ast_channel_hangup_request_type(), sub_channel_blob_handler, NULL);
if (r) {
return AST_MODULE_LOAD_FAILURE;
}
bridge_router = stasis_message_router_create(ast_bridge_topic_all_cached());
if (!bridge_router) {
return AST_MODULE_LOAD_FAILURE;
}
r |= stasis_message_router_add(bridge_router, stasis_cache_update_type(), sub_bridge_snapshot_handler, NULL);
r |= stasis_message_router_add(bridge_router, ast_bridge_merge_message_type(), sub_bridge_merge_handler, NULL);
r |= stasis_message_router_add(bridge_router, ast_channel_entered_bridge_type(), sub_bridge_enter_handler, NULL);
r |= stasis_message_router_add(bridge_router, ast_channel_left_bridge_type(), sub_bridge_leave_handler, NULL);
if (r) {
return AST_MODULE_LOAD_FAILURE;
}
return AST_MODULE_LOAD_SUCCESS;
}
static int unload_module(void)
{
int r = 0;
stasis_message_router_unsubscribe_and_join(channel_router);
channel_router = NULL;
stasis_message_router_unsubscribe_and_join(bridge_router);
bridge_router = NULL;
ao2_cleanup(apps_registry);
apps_registry = NULL;
ao2_cleanup(app_controls);
app_controls = NULL;
ao2_cleanup(app_bridges);
app_bridges = NULL;
return r;
}
AST_MODULE_INFO(ASTERISK_GPL_KEY, AST_MODFLAG_GLOBAL_SYMBOLS, "Stasis application support",
.load = load_module,
.unload = unload_module,
);