We have historically cached two types of values
on a per-request basis inside of memory:
* linkifiers
* display recipients
Both of these caches were hand-written, and they
both actually cache values that are also in memcached,
so the per-request cache essentially only saves us
from a few memcached hits.
I think the linkifier per-request cache is a necessary
evil. It's an important part of message rendering, and
it's not super easy to structure the code to just get
a single value up front and pass it down the stack.
I'm not so sure we even need the display recipient
per-request cache any more, as we are generally pretty
smart now about hydrating recipient data in terms of
how the code is organized. But I haven't done thorough
research on that hypotheseis.
Fortunately, it's not rocket science to just write
a glorified memoize decorator and tie it into key
places in the code:
* middleware
* tests (e.g. asserting db counts)
* queue processors
That's what I did in this commit.
This commit definitely reduces the amount of code
to maintain. I think it also gets us closer to
possibly phasing out this whole technique, but that
effort is beyond the scope of this PR. We could
add some instrumentation to the decorator to see
how often we get a non-trivial number of saved
round trips to memcached.
Note that when we flush linkifiers, we just use
a big hammer and flush the entire per-request
cache for linkifiers, since there is only ever
one realm in the cache.
We generally want to avoid extra moving parts when we
stringify objects. We also want to phase out the use
of get_display_recipient for streams.
Note that we still hit get_display_recipient to
stringify DM and huddle objects, and it's kind of ugly
how we do it, but that's outside the scope of my
current PR.
This commit renames the 'PRIVATE_MESSAGE' attribute of the
'NotificationTriggers' class to 'DIRECT_MESSAGE'.
Custom migration to update the existing value in the database.
It includes 'TODO/compatibility' code to support the old
notification trigger value 'private_message' in the
push notification queue during the Zulip server upgrades.
Earlier 'private_message' was one of the possible values for the
'trigger' property of the '[`POST /zulip-outgoing-webhook`]' response;
Update the docs to reflect the change in the above-mentioned trigger
value.
This commit adds code to pass all the required arguments to
select_related call for Message objects such that only the
required related fields are fetched from the database.
Previously, we did not pass any arguments to select_related,
so all the directly and indirectly related fields were fetched
when many of them were actually not being used and made the
query unnecessarily complex.
This commit updates the code to pass "realm" and "recipient" as
arguments to select_related call in get_stream_by_id_in_realm.
Previously, since there was no arguments, it fetched
can_remove_subscribers_group and the related fields of
"Realm" model as well which were not being used, but
did not fetch "recipient" as it is a nullable field.
This commit removes get_huddle_recipient function and we now use
get_or_create_huddle in get_recipient_from_user_profiles.
As a result of this change, we do not fetch the recipient from
Huddle object but instead get it using the "id" and "recipient_id"
fields available from Huddle object like we do for a personal
message. This change allows us to not fetch recipient object
using select_related when querying the Huddle object.
We now fetch recipient object when querying "Huddle" object in
get_or_create_huddle_backend as this query is eventually used
to get the recipient object only in get_huddle_recipient.
This commit also updates the select_related call in the code to
populate Huddle objects in cache to pass "Recipient" as argument.
Previously no argument was passed to select_related and thus no
related objects were being fetched, with no non-null related fields
being present.
This commit adds code to pass stream traffic data using
the "stream_weekly_traffic" field in stream objects.
We already include the traffic data in Subscription objects,
but the traffic data does not depend on the user to stream
relationship and is stream-only information, so it's better
to include it in Stream objects. We may remove the traffic
data and other stream information fields for Subscription
objects in future.
This will help clients to correctly display the stream
traffic data in case where client receives a stream
creation event and no subscription event, for an already
existing stream which the user did not have access to before.
This commit changes the code to not use get_client_data
function and instead use `stream_to_dict` function to
get the stream data in a dictionary form. This is a
prep commit add stream traffic data to Stream objects.
Previously, this code:
```python3
old_archived_attachments = ArchivedAttachment.objects.annotate(
has_other_messages=Exists(
Attachment.objects.filter(id=OuterRef("id"))
.exclude(messages=None)
.exclude(scheduled_messages=None)
)
).filter(messages=None, create_time__lt=delta_weeks_ago, has_other_messages=False)
```
...protected from removal any ArchivedAttachment objects where there
was an Attachment which had _both_ a message _and_ a scheduled
message, instead of _either_ a message _or_ a scheduled message.
Since files are removed from disk when the ArchivedAttachment rows are
deleted, this meant that if an upload was referenced in two messages,
and one was deleted, the file was permanently deleted when the
ArchivedMessage and ArchivedAttachment were cleaned up, despite being
still referenced in live Messages and Attachments.
Switch from `.exclude(messages=None).exclude(scheduled_messages=None)`
to `.exclude(messages=None, scheduled_messages=None)` which "OR"s
those conditions appropriately.
Pull the relevant test into its own file, and expand it significantly
to cover this, and other, corner cases.
Earlier the API endpoints related to streams accepts and returns a
field `can_remove_subscribers_group_id` which represents the ID
of user_group whose members can remove subscribers from stream.
This commit renames this field to `can_remove_subscribers_group`.
Dropping support for url_prefix for RealmPlayground, the server now uses
url_template instead only for playground creation, retrieval and audit
logging upon removal.
This does the necessary handling so that url_template is expanded with
the extracted code.
Fixes#25723.
Signed-off-by: Zixuan James Li <p359101898@gmail.com>
As an intermediate step before we fully support url_template for realm
playgrounds, we populate url_template in the backend ensuring that all
the new entries will be validated. With a later backfilling migration,
we prepare the database such that all the records will have a valid URL
template.
Signed-off-by: Zixuan James Li <p359101898@gmail.com>
Updates the realm field `default_code_block_language` to have a default
value of an empty string instead of None. Also updates the web-app to
check for the empty string and not `null` to indicate no default is set.
This means that both new realms and existing realms that have no default
set will have the same value for this setting: an empty string.
Previously, new realms would have None if no default was set, while realms
that had set and then unset a value for this field would have an empty
string when no default was set.
We used to access the complete objects for UserProfile foreign
keys like "bot_owner" and "default_sending_stream", where we only
needed ID of them.
This commit fixes some of such instances and now we directly get
the id using "bot_owner_id" and "default_sending_stream_id" so
that we can avoid the unnecessary complexity of accessing the
complete object.
This commit updates the select_related calls in queries to get
UserProfile objects in get_user, get_user_by_delivery_email,
get_user_profile_by_id, get_user_profile_by_id_in_realm and
get_user_profile_by_api_key functions to pass "realm" and
"bot_owner" as arguments to select_related call.
These functions are used in different parts of code to get
the UserProfile object and realm is accessed using the user
object at many places.
"bot_owner" field is also used in some places like to check
whether a bot can access a stream, to check whether a user
can change modify another user, in webhooks code to send the
message to the bot owner, and in tests as well. There can be
some places where the bot owner is not required and in most
such cases the code would only be accessed for human users,
which means the bot_owner will be null for these cases and
would avoid complexity and performance issues.
Note that previously, no arguments were passed to select_related
and thus only realm field was fetched during the query.
This commit updates the select_related calls in queries to
get UserProfile object in get_syste_bot function pass "realm"
as argument to select_related call.
The "get_system_bot" call function is mostly used to get cross
realm bot which are used as senders to send messages.
The fields like default_events_register_stream and recipient
are not required for these cases. The bot_owner field is used
to check access to a stream to send message but the cross-realm
bots are handled differently and the bot_owner check is not
required.
Also, note that "realm" is the only non-null foreign key field
in UserProfile object, so select_related() was only fetching
realm object previously as well. But we should still pass
"realm" as argument in select_related call so that we can make
sure that only required fields are selected in case we add
more foreign keys to UserProfile in future.
This commit updates select_related call in get_user_profile_by_email
to pass "realm" as argument.
This function is intended to be used for manual manage.py shell
work so we just keep the behavior same as before as "realm" is
the only non-null related field in UserProfile.
We do not use any related fields for the UserProfile objects
fetched by get_active_users, so we can simply remove the
select_related call.
The user object from get_active_users was used to get realm
but since get_active_users called from a realm object we can
directly use that realm object. This change also leads to
some changes in the cache code where we now pass the realm
to the function instead of selecting it from UserProfile object.
This code removes a lot of complexity with very likely
positive overall impact on system performance and
negligible downside.
We already cache display recipients on a per-user
level, so there's no need for another cache layer on
top of that that keys them with recipient ids.
We avoid strange things where Alice/Bob and Bob/Charlie
get put into the top layer cache and then we still have
a cache miss on Alice/Charlie despite the lower level
cache being able to support per-user lookups.
This change does introduce an extra database round trip
if any of our messages have a huddle, but the query is
extremely cheap, and we can always try to cache that
function more directly or try to re-use some of our
other huddle-based caches.
As part of this, we clean up the names for the
lower-level per-user cache of display recipients, and
we simplify the cache keys.
We also stop passing in a full Recipient object to the
`bulk_get_huddle_user_ids` functions.
The local impact of this change should be easy to
measure (at least approximately), since we use this
function every time a user gets messages via the
/messages endpoint.
We restrict the columns, avoid quadratic looping,
and don't bother with order_by.
We also return the user ids (per recipient) as
sets, since that's how the only caller uses the
info (albeit implicitly via set.union accepting
a list).
It’s unclear what was supposed to be “safe” about this wrapper. The
hashlib API is fine without it, and we don’t want to encourage further
use of SHA-1.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Translators benefit from the extra information in the field names, and
need the reordering freedom that isn’t available with multiple
positional fields.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
This commit completes the notifications part of the @topic
wildcard mention feature.
Notifications are sent to the topic participants for the
@topic wildcard mention.
The active realm emoji are just a subset of all your
realm emoji, so just use a single cache entry per
realm.
Cache misses should be very infrequent per realm.
If a realm has lots of deactivated realm emoji, then
there's a minor expense to deserialize them, but that
is gonna be dwarfed by all the other more expensive
operations in message-send.
I also renamed the two related functions. I erred on
the side of using somewhat verbose names, as we don't
want folks to confuse the two use cases. Fortunately
there are somewhat natural affordances to use one or
the other, and mypy helps too.
Finally, I use realm_id instead of realm in places
where we don't need the full Realm object.
This commit removes "@" from name of role-based system groups
since we have added a restricion on having user group names
starting with "@" in the previous commit as they look odd in
mention syntax.
We also add a migration in this commit to update the name of
role-based system groups in existing realms to remove "@"
from the name. This migration also updates the names of
non-system user groups by removing the invalid prefixes
from their names and if there is a group already with that
name, we insted name the group as "group:{group_id}".
Fixes#26148.
We do not allow user group names to start with "@", "role:",
"user:", "stream:" and "channel:".
Group names starting with "@" look odd in mentions and
"role:", "user:" and "stream:" prefixes are reserved for
system groups which will be used in the new groups-based
permission model. We do not allow "channel:" prefix for
now just to be safe in a case where we use it instead of
"stream:" prefix for stream based groups in future.
Fixes part of #26148.
Previously we had database level restriction on length of
user group names. Now we add the same restriction to API
level as well, so we can return a better error response.
We remove the cache functionality for the
get_realm_stream function, and we also change it to
return a thin Stream object (instead of calling
select_related with no arguments).
The main goal here is to remove code complexity, as we
have been prone to at least one caching validation bug
related to how Realm and UserGroup interact. That
particular bug was more theoretical than practical in
terms of its impact, to be clear.
Even if we were to be perfectly disciplined about only
caching thin stream objects and always making sure to
delete cache entries when stream data changed, we would
still be prone to ugly situations like having
transactions get rolled back before we delete the cache
entry. The do_deactivate_stream is a perfect example of
where we have to consider the best time to unset the
cache. If you unset it too early, then you are prone to
races where somebody else churns the cache right before
you update the database. If you set it too late, then
you can have an invalid entry after a rollback or
deadlock situation. If you just eliminate the cache as
a moving part, that whole debate is moot.
As the lack of test changes here indicates, we rarely
fetch streams by name any more in critical sections of
our code.
The one place where we fetch by name is in loading the
home page, but that is **only** when you specify a
stream name. And, of course, that only causes about an
extra millisecond of time.
This changes bulk_get_streams so that it just uses the
database all the time. Also, we avoid calling
select_related(), so that we just get back thin and
tidy Stream objects with simple queries.
About not caching any more:
It's actually pretty rare that we fetch streams by name
in the main application. It's usually API requests that
send in stream names to find more info about streams.
It also turns out that for large queries (>= ~30 rows
for my measurements) it's more efficent to hit the
database than memcached. The database is super fast at
scale; it's just the startup cost of having Django
construct the query, and then having the database do
query planning or whatever, that slows us down. I don't
know the exact bottleneck, but you can clearly measure
that one-row queries are slow (on the order of a full
millisecond or so) but the marginal cost of additional
rows is minimal assuming you have a decent index (20
microseconds per row on my droplet).
All the query-count changes in the tests revolve around
unsubscribing somebody from a stream, and that's a
particularly odd use case for bulk_get_streams, since
you generally unsubscribe from a single stream at a
time. If there are some use cases where you do want to
unsubscribe from multiple streams, we should move
toward passing in stream ids, at least from the
application. And even if we don't do that, our cost for
most queries is a couple milliseconds.
This add audit log entries when any group based setting of a user group
is updated. We store both the old and new values in extra_data, along
with the name of that setting. Entries populated during user group creation
are hardcoded to track "can_mention_group".
Potentially we can adjust "set_defaults_for_group_settings" so that it
populates realm audit logs with it, but that is out of scope for this change.
We use an atomic transaction so that the audit logs are committed
together with the updates.
Signed-off-by: Zixuan James Li <p359101898@gmail.com>
This prep commit replaces the 'wildcard' keyword in the codebase
with 'stream_wildcard' at some places for better readability, as
we plan to introduce 'topic_wildcards' as a part of the
'@topic mention' project.
Currently, 'wildcards = ["all", "everyone", "stream"]' which is an
alias to mention everyone in the stream, hence better renamed as
'stream_wildcards'.
Eventually, we will have:
'stream_wildcard' as an alias to mention everyone in the stream.
'topic_wildcard' as an alias to mention everyone in the topic.
'wildcard' refers to 'stream_wildcard' and 'topic_wildcard' as a whole.
Subsequent commits will add "on_delete=models.RESTRICT"
relationships, which will result in the AlertWord
objects being deleted after Realm has been deleted from
the database.
In order to handle this, we update realm_alert_words_cache_key,
realm_alert_words_automaton_cache_key, and flush_realm_alert_words
functions to accept realm_id as parameter instead of realm
object, so that the code for flushing the cache works even
after the realm is deleted. This change is fine because
eventually only realm_id is used by these functions and there
is no need of the complete realm object.
Subsequent commits will add "on_delete=models.RESTRICT"
relationships, which will result in the Attachment
objects being deleted after Realm has been deleted from
the database.
In order to handle this, we update
get_realm_used_upload_space_cache_key function to accept
realm_id as parameter instead of realm object, so that
the code for flushing the cache works even after the
realm is deleted. This change is fine because eventually
only realm_id is used by this function and there is no
need of the complete realm object.
Subsequent commits will add "on_delete=models.RESTRICT"
relationships, which will result in the UserProfile
objects being deleted after Realm has been deleted from
the database.
In order to handle this, we update bot_dicts_in_realm_cache_key
function to accept realm_id as parameter instead of realm
object, so that the code for flushing the cache works even
after the realm is deleted. This change is fine because
eventually only realm_id is used by this function and there is
no need of the complete realm object.
Subsequent commits will add "on_delete=models.RESTRICT"
relationships, which will result in the RealmEmoji
objects being deleted after Realm has been deleted from
the database.
In order to handle this, we update get_realm_emoji_dicts,
get_realm_emoji_cache_key, get_active_realm_emoji_cache_key,
get_realm_emoji_uncached and get_active_realm_emoji_uncached
functions to accept realm_id as parameter instead of realm
object, so that the code for flushing the cache works even
after the realm is deleted. This change is fine because
eventually only realm_id is used by these functions and
there is no need of the complete realm object.
This commit adds default_group_name field to GroupPermissionSetting
type which will be used to store the name of the default group for
that setting which would in most cases be one of the role-based
system groups. This will be helpful when we would have multiple
settings and we would need to set the defaults while creating
realm and streams.
This commit makes it possible for users to control the
audible desktop notifications for messages sent to followed topics
via a global notification setting.
There is no support for configuring this setting through the UI yet.
This commit makes it possible for users to control the
visual desktop notifications for messages sent to followed topics
via a global notification setting.
There is no support for configuring this setting through the UI yet.
Steve Howell
Prakhar Pratyush
Sahil Batra
Anders Kaseorg
Alex Vandiver
Ujjawal Modi
Zixuan James Li
Lauryn Menard