This implements a significant performance optimization for users
clicking the `Private messages` narrow in the Zulip UI, especially for
those users who do not have 50 recent private messages in an
organization with a lot of stream message traffic (because then
previously, postgres needed to scan through a huge amount of history
to find enough private messages).
The database index powering it can also support many other queries we
might want to do in the future to support "recent conversations" type
features.
Fixes#6896.
Since otp_encrypt_api_key only encrypts API keys, it doesn't require
access to the full UserProfile object to work properly. Now the
parameter it accepts is just the API key.
This is preparatory refactoring for removing the api_key field on
UserProfile.
random_api_key, the function we use to generate random tokens for API
keys, has been moved to zerver/lib/utils.py because it's used in more
parts of the codebase (apart from user creation), and having it in
zerver/lib/create_user.py was prone to cyclic dependencies.
The function has also been renamed to generate_api_key to have an
imperative name, that makes clearer what it does.
Now reading API keys from a user is done with the get_api_key wrapper
method, rather than directly fetching it from the user object.
Also, every place where an action should be done for each API key is now
using get_all_api_keys. This method returns for the moment a single-item
list, containing the specified user's API key.
This commit is the first step towards allowing users have multiple API
keys.
Importing the Django test client is somewhat expensive, and we only
use it within one view function that's not used in production. So
there's a significant startup-time performance optimization in doing
an import inside the view code.
The use_first_unread_anchor parameter allows automatically setting the
anchor to the first message that hasn't been read in this narrow.
Therefore it isn't necessary to specify an anchor when this parameter is
enabled.
Note from Tim: Arguably, we should think about making
`use_first_unread_anchor` the default behavior when anchor is
unspecified, but that's for later consideration.
This fixes a bug where administrators couldn't remove private
unsubscribed streams from the "default streams" list, because
access_stream_by_name didn't give them access to the stream object.
This renames Realm.restricted_to_domain field to
emails_restricted_to_domains, for greater clarity as to what it does
just from seeing the setting name, without having to look it up.
Fixes part of #10042.
It's sorta an unusual state to get into, to have a user own a
deactivated bot, when they can't create a bot of that type, but
definitely a valid possibility that we should be checking for.
Fixes#10087.
This setting isn't intended to exist long term, but instead to make it
possible to merge our search pills code before we're ready to cut over
production environments to use it.
Various pieces of our thumbor-based thumbnailing system were already
merged; this adds the remaining pieces required for it to work:
* a THUMBOR_URL Django setting that controls whether thumbor is
enabled on the Zulip server (and if so, where thumbor is hosted).
* Replaces the overly complicated prototype cryptography logic
* Adds a /thumbnail endpoint (supported both on web and mobile) for
accessing thumbnails in messages, designed to support hosting both
external URLs as well as uploaded files (and applying Zulip's
security model for access to thumbnails of uploaded files).
* Modifies bugdown to, when THUMBOR_URL is set, render images with the
`src` attribute pointing /thumbnail (to provide a small thumbnail
for the image), along with adding a "data-original" attribute that
can be used to access the "original/full" size version of the image.
There are a few things that don't work quite yet:
* The S3 backend support is incomplete and doesn't work yet.
* The error pages for unauthorized access are ugly.
* We might want to rename data-original and /thumbnail?size=original
to use some other name, like "full", that better reflects the fact
that we're potentially not serving the original image URL.
In this commit we add a new endpoint so as to have a way of fetching
topic history for a given stream id without having to be logged in.
This can only happen if the said stream is web public otherwise we
just return an empty topics list. This endpoint is quite analogous
to get_topics_backend which is used by our main web app.
In this commit we also do a bit of duplication regarding the query
responsible for fetching all the topics from DB. Basically this
query is exactly the same as what we have in the
get_topic_history_for_stream function in actions.py. Basically
duplicating now is the right thing to do because this query is
really gonna change when we add another criteria for filtering
messages which is:
Only topics for messages which were sent during the period the
corresponding stream was web public should be returned.
Now when we will do this, the query will change and thus it won't
really be a code duplication!
POST and DELETE operations in /users/me/alert_words may leave the
user's list of alert words in an unknown state: POSTing adds words to a
list that the client may not know from the begining, and the same with
DELETE.
Replying with the current status of the alert words list is the best way
of letting the client alter the list and knowing its contents after
being updated with a single query.
This is especially useful taking into account that POSTing words that
were already present and DELETing non-existing words both produce a
successful response.
An extra test has been added to avoid leaving GET /users/me/alert_words
too untested.
Querying an endpoint with no information (thus a noop) and it producing
a successful response doesn't seem to be expected.
Given the case that the client makes such query with no content it will
probably be unintentional and the API should let them know about it.
This is all the plumbing that makes it possible to enable the
stream_email_notifications setting via the Zulip API. The flag still
doesn't do anything yet, but this is a nice checkpoint along the way
to implementing this feature.
Given that we allow adding emoji reactions by only using the
emoji_name, we should offer the same possibility for removing
reactions to make the experience for API clients not require looking
up emoji codes.
Since this is an additional optional parameter, this also preserves
backward compatibility.
Complete, correct implementations of Zulip's emoji reactions API need
to send both emoji_code and emoji_name in order to add a reaction;
this is important for corner cases around clicking on a reaction in a
message that was first reacted to a year ago, when the emoji
name->code mappings have changed for the given code point in the
intervening time.
However, for folks building tools using the Zulip API, that corner
case is not particularly common; as a result, it makes sense to offer
an interface that allows adding a reaction by only specifying the
emoji name.
This is why the only field that needs to be required is emoji_name,
which can now be mapped to a single emoji. Both fields will be
necessary when "voting" an old reaction, but since we stil allow
specifying the two of them, these changes offer retrocompatibility.
This adds a new settings, SOCIAL_AUTH_SUBDOMAIN, which specifies which
domain should be used for GitHub auth and other python-social-auth
backends.
If one is running a single-realm Zulip server like chat.zulip.org, one
doesn't need to use this setting, but for multi-realm servers using
social auth, this fixes an annoying bug where the session cookie that
python-social-auth sets early in the auth process on the root domain
ends up masking the session cookie that would have been used to
determine a user is logged in. The end result was that logging in
with GitHub on one domain on a multi-realm server like zulipchat.com
would appear to log you out from all the others!
We fix this by moving python-social-auth to a separate subdomain.
Fixes: #9847.
The only changes visible at the AST level, checked using
https://github.com/asottile/astpretty, are
zerver/lib/test_fixtures.py:
'\x1b\\[(1|0)m' ↦ '\\x1b\\[(1|0)m'
'\\[[X| ]\\] (\\d+_.+)\n' ↦ '\\[[X| ]\\] (\\d+_.+)\\n'
which is fine because re treats '\\x1b' and '\\n' the same way as
'\x1b' and '\n'.
Signed-off-by: Anders Kaseorg <andersk@mit.edu>
When GETting an unedited message's edit history, the server wasn't able
to reply properly and produced a 500 error.
Now when that happens, we return a message history that only contains
the original message.
We're adding more stream types, e.g. splitting private streams into
with/without shared history, adding publicly-archived streams, adding
announce-only streams, etc. So maintaining this text is going to get more
complicated over time.
Also, the right place to explain this stuff is in the stream header, or near
the z-in-a-circle.
This commit also adds translation tags to the messages.
Move the zcommands from '/views/messages.py' to
'/lib/zcommand'.
Also, move the zcommand tests from '/tests/test_messages.py'
to '/tests/test_zcommand'.
This adds a common function `access_user_by_id` to access user id
within same realm, complete with a full suite of unit tests.
Tweaked by tabbott to make the test much more readable.
These two slash commands now use zcommand to talk to
the server, so we have no Message overhead, and if you're
on a stream, you no longer spam people by accident.
The commands now also give reasonable messages
if you are already in the mode you ask for.
It should be noted that by moving these commands out of
widget.py, they are no longer behind the ALLOW_SUB_MESSAGES
setting guard.
This adds a /ping command that will be useful for users
to see what the round trip to the Zulip server is (including
only a tiny bit of actual server time to basically give a
200).
It also introduce the "/zcommand" endpoint and zcommand.js
module.
For some reason in my original version I was sending both
content and data to the client for submessage events,
where data === JSON.parse(content). There's no reason
to not just let the client parse it, since the client
already does it for data that comes on the original
message, and since we might eventually have non-JSON
payloads.
The server still continues to validate that the payload
is JSON, and the client will blueslip if the server
regressses and sends bad JSON for some reason.
We now have a simple algorithm: First, look at the URL path
(e.g. /de/, which is intended to be an override). Second, look at the
language the user has specified in their settings.
This adds a common function `access_bot_by_id` to access bot id within
same realm. It probably fixes some corner case bugs where we weren't
checking for deactivated bots when regenerating API keys.
Previously, if you had LDAPAuthBackend enabled, we basically blocked
any other auth backends from working at all, by requiring the user's
login flow include verifying the user's LDAP password.
We still want to enforce that in the case that the account email
matches LDAP_APPEND_DOMAIN, but there's a reasonable corner case:
Having effectively guest users from outside the LDAP domain.
We don't want to allow creating a Zulip-level password for a user
inside the LDAP domain, so we still verify the LDAP password in that
flow, but if the email is allowed to register (due to invite or
whatever) but is outside the LDAP domain for the organization, we
allow it to create an account and set a password.
For the moment, this solution only covers EmailAuthBackend. It's
likely that just extending the list of other backends we check for in
the new conditional on `email_auth_backend` would be correct, but we
haven't done any testing for those cases, and with auth code paths,
it's better to disallow than allow untested code paths.
Fixes#9422.
Previously, if both EmailAuthBackend and LDAPAuthBackend were enabled,
LDAP users could set a password using EmailAuthBackend and continue to
use that password, even if their LDAP account was later deactivated.
That configuration wasn't supported at all before, so this doesn't fix
a pre-existing security issue, but now that we're making that a valid
configuration, we need to cover this case.
This should make it easier for us to iterate on a less-dense Zulip.
We create two classes on body, less_dense_mode and more_dense_mode, so
that it's easy as we refactor to separate the two concepts from things
like colors that are independent.
API users, particularly bots, can now send a field
called "widget_content" that will be turned into
a submessage for the web app to look at. (Other
clients can still rely on "content" to be there,
although it's up to the bot author to make the
experience good for those clients as well.)
Right now widget_content will be a JSON string that
encodes a "zform" widget with "choices." Our first
example will be a trivia bot, where users will see
something like this:
Which fruit is orange in color?
[A] orange
[B] blackberry
[C] strawberry
The letters will be turned into buttons on the webapp
and have canned replies.
This commit has a few parts:
- receive widget_content in the request (simply
validating that it's a string)
- parse the JSON in check_message and deeply
validate its structure
- turn it into a submessage in widget.py
This commit adds a view which will be used to process login requests,
adds an AuthenticationTokenForm so that we can use TextField widget for
tokens, and activates two factor authentication code path whenever user
tries to login.
This should significantly improve the user experience for creating
additional accounts on zulipchat.com.
Currently, disabled in production pending some work on visual styling.
This query was incorreclty not checking whether a user was deactivated
before managing their subscriptions.
This isn't an important bug, but should prevent some weird corner
cases (like trying to send a notification PM to a deactivated user,
which fails).
We send add events on upload, update events when sending a message
referencing it, and delete updates on removal.
This should make it possible to do real-time sync for the attachments
UI.
Based in part on work by Aastha Gupta.
These decorators will be part of the process for disabling access to
various features for guest users.
Adding this decorator to the subscribe endpoint breaks the guest users
test we'd just added for the subscribe code path; we address this by
adding a more base-level test on filter_stream_authorization.
This removes a check on invite_only, that should have been a check on
history_public_to_subscribers. In addition to fixing a bug for zephyr
realms, it also makes "more topics" work correctly for realms using
the new settings for stream history being public to subscribers.
We haven't seen significant traffic from the legacy desktop app in
over a year, and users using it get a warning to upgrade since last
summer, so it's probably OK to stop providing special fonts for it.
This commit adds a new field history_public_to_subscribers to the
Stream model, which serves a similar function to the old
settings.PRIVATE_STREAM_HISTORY_FOR_SUBSCRIBERS; we still use that
setting as the default value for new streams to avoid breaking
backwards-compatibility for those users before we are ready with an
actual UI for users to choose directly.
This also comes with a migration to set the value of the new field for
existing streams with an algorithm matching that used at runtime.
With significant changes by Tim Abbott.
This is an initial part of our efforts on #9232.
The handlebars error message is just for the manual development
environment; this prevents the state of compiling handlebars templates
from run-dev.py from potentially causing the unit tests to fail.
Add realm setting to set time limit for message deleitng.
Set default value of message_content_delete_limit_seconds
to 600 seconds(10 min).
Thanks to Shubham Dhama for rebasing and reworking this. Some final
edits also done by Tim Abbott.
Fixes#7344.
This should make it easier to find the templates that are actually
part of the core webapp, instead of having them all mixed together
with the portico pages.
The main change here is to send a proper confirmation link to the
frontend in the `confirm_continue_registration` code path even if the
user didn't request signup, so that we don't need to re-authenticate
the user's control over their email address in that flow.
This also lets us delete some now-unnecessary code: The
`invalid_email` case is now handled by HomepageForm.is_valid(), which
has nice error handling, so we no longer need logic in the context
computation or template for `confirm_continue_registration` for the
corner case where the user somehow has an invalid email address
authenticated.
We split one GitHub auth backend test to now cover both corner cases
(invalid email for realm, and valid email for realm), and rewrite the
Google auth test for this code path as well.
Fixes#5895.
By moving all of the logic related to the is_signup flag into
maybe_send_to_registration, we make the login_or_register_remote_user
function quite clean and readable.
The next step is to make maybe_send_to_registration less of a
disaster.
The code in maybe_send_to_registration incorrectly used the
`get_realm_from_request` function to fetch the subdomain. This usage
was incorrect in a way that should have been irrelevant, because that
function only differs if there's a logged-in user, and in this code
path, a user is never logged in (it's the code path for logged-out
users trying to sign up).
This this bug could confuse unit tests that might run with a logged-in
client session. This made it possible for several of our GitHub auth
tests to have a totally invalid subdomain value (the root domain).
Fixing that bug in the tests, in turn, let us delete a code path in
the GitHub auth backend logic in `backends.py` that is impossible in
production, and had just been left around for these broken tests.
This code path has actually been dead for a while (since
`invalid_subdomain` gets set to True only when `user_profile` is
`None`). We might want to re-introduce it later, but for now, we
eliminate it and the artificial test that provided it with test
coverage.
This is done mainly because this backend has the simplest code path
for calling login_or_register_remote_user, more than because we expect
this case to come up. It'll make it easier to write unit tests for
the `invalid_subdomain` corner case.
This is a mobile-specific endpoint used for logging into a dev server.
On mobile without this realm_uri it's impossible to send a login request
to the corresponding realm on the dev server and proceed further; we can
only guess, which doesn't work for using multiple realms.
Also rename the endpoint to reflect the additional data.
Testing Plan:
Sent a request to the endpoint, and inspected the result.
[greg: renamed function to match, squashed renames with data change,
and adjusted commit message.]
After some thinking, I don't think there's any actual value to doing
the ../ style relative links here, whereas there is actual harm from
the links being slightly broken in the current model. We fix this by
just using /#settings as the URL.
Fixes#8978.
For certain queries where both include_history and
use_first_unread_anchor are set to True, we were excluding
historical rows. Now we only use the use_first_unread_anchor
flag to filter rows that we use to find the anchor, without
having it filter the actual search results.
The bug went unreported for a long time, because it only
affected mobile users who had newly subscribed to streams.
Note that we make a small change to the test called
test_use_first_unread_anchor_with_muted_topics, which has
a very scary comment about being "arcane" and "be
absolutely sure you know what you're doing." I think it's
fine.
Also, the new test code would fail before this fix, so it
should help prevent future regressions.
Fixes#8958
This is a bit more than a pure refactor, because we duplicate a
chunk of code to calculate a query inside of
find_first_unread_anchor(), so we're doing a bit more work
than before.
We need this refactoring to start decoupling find_first_unread_anchor
from get_messages_backend for the case where include_history is
True. This will happen in a subsequent commit.
The only test that changes here is a direct test on
find_first_unread_anchor(). All other tests pass without
modification, and we have decent coverage on get_messages_backend.
We use an array now to build up the list of search operands and
then consolidate the special search handling after the loop (which
means setting the flag, putting two more columns in the query, and
using ' '.join to build the string).
We have a debugging statement for some obscure errors we get
when narrows have search terms. We now show all the narrow
operators. This isn't really to improve debugging; it's more
to make it easier in the next commit to extract a function
that would make search_term have to be passed back in a tuple.
But it shouldn't hurt debugging either.
Refactoring in this file had resulted in the logic for
html_settings_link being duplicated and extra logic being needed to
ensure these variables were set where they were needed.
This fixes subscriptions_html not being rendered properly in the /help
and /api pages, in addition to removing duplicate code.
This is a pure refactoring and just pulls the function out
to the top level of the module. (The prior commit extracted
it inside a larger function to make a nicer diff.)
The new name can_access_stream_history_by_name gets to the point of
what this function actually does. And passing in a user object lets
us define what this does based on the user subscribed.
This fixes a bug where the endpoint for editing bot users would allow
an organization administrator to edit the full name of a bot user.
A combination of this an another recently fixed bug made it possible
for this process to set a `bot_owner` for a non-bot user; so we also
include a migration to fix that for any users that might have had our
model invariants corrupted in that way.
This bot was basically a duplicate of NOTIFICATION_BOT for some
specific corner cases, and didn't add much value. It's better to just
eliminate it, which also removes some ugly corner cases around what
happens if the user account doesn't exist.
Add function in user-groups.py for getting member ids
for a group.
Update view to enforce checks for modifying user-groups.
Only admins and user group members can modify user-groups.
Applies the logic to allow community members to edit topics
of others' messages if this setting is True. Otherwise,
only administrators can update the topic of others' messages.
This logic includes a 24-hour time limit for community topic editing.
This commit asserts that parse_user_agent never returns None. The
RegEx will match any string, so that `match` is never None. This
brings test coverage of lib/user_agent.py to 100%. Changes were also
made in test/test_decorators.py and views/compatibility.py to reflect
that parse_user_agent cannot return None.
Improves: #7089.
Fixes: #8779.
It's possible that this won't work with some versions of the
third-party backend, but tabbott has tested carefully that it does
work correctly with the Apache basic auth backend in our test
environment.
In this commit we start to support redirects to urls supplied as a
'next' param for the following two backends:
* GoogleOAuth2 based backend.
* GitHubAuthBackend.
This commit migrates realm emoji to be addressed by their `id` rather
than their name. This fixes a long standing issue which was causing
an error on uploading an emoji with same name as a deactivated realm
emoji.
Fixes: #6977.
We now consistently set our query limits so that we get at
least `num_after` rows such that id > anchor. (Obviously, the
caveat is that if there aren't enough rows that fulfill the
query, we'll return the full set of rows, but that may be less
than `num_after`.) Likewise for `num_before`.
Before this change, we would sometimes return one too few rows
for narrow queries.
Now, we're still a bit broken, but in a more consistent way. If
we have a query that does not match the anchor row (which could
be true even for a non-narrow query), but which does match lots
of rows after the anchor, we'll return `num_after + 1` rows
on the right hand side, whether or not the query has narrow
parameters.
The off-by-one semantics here have probably been moot all along,
since our windows are approximate to begin with. If we set
num_after to 100, its just a rough performance optimization to
begin with, so it doesn't matter whether we return 99 or 101 rows,
as long as we set the anchor correctly on the subsequent query.
We will make the results more rigorous in a follow up commit.
We start to force downloads for the attachment files. We do this
for all files except images or pdf's. We would like images or pdf's
to open up in browser itself.
Tweaked by tabbott for comment clarity and correctness.
This will allow realm admins to remove others from private stream to
which the realm administrator is not subscribed; this is important for
managing those streams, because previously nobody could remove users
from private streams that didn't have any realm administrators
subscribed.
This will allow realm admins to access subscribers of unsubscribed
private stream. This is a preparatory commit for letting realm admins
remove those users.
This will allow realm admins to update the names and descriptions of
private streams even if they are not subscribed, which fixes the buggy
behavior that previously nobody could(!).
This generic function isolates the before/after logic that really
is independent of Message and doesn't need to clutter up
`get_messages_backend`. Also, introducing a new namespace
reduces some shadowing/mutation with variables like `query`.
It's a pure code move, with some very minor renaming (e.g.
inner_msg_id_col -> id_col).
If anchor is 0, there is no sense doing a before_query.
Likewise, if anchor is `LARGER_THAN_MAX_MESSAGE_ID`, there is
no sense doing an after_query.
We introduce variables called `need_before_query` and
`need_after_query` to enforce those conditions.
This also adds some comments explaining the fallthrough case
where neither query makes sense.
If use_first_unread_anchor is set and we don't have any unread
messages, then our anchor is effectively "positive infinity" and
we can streamline queries.
In the past we'd have clauses like `message_id <= 999999999999999`
in the query that were harmless but crufty.
We want to say `if num_after > 0` when we expect num_after to be
a positive integer. We don't want any confusion that we will
execute the blocks for values of -7 or None.
Apparently, we did essentially all the work to support showing full
topic history to newly subscribed users from a data flow perspective,
but didn't actually enable this feature by having the topic history
endpoint grant access to historical topics. This fixes that gap.
I'm not altogether happy with how the code and tests read for this
feature; the code itself has more duplication than I'd like, and the
tests do too, but it works.
"incorrect" here means rejected by a bot's validate_config() method.
A common scenario for this is validating API keys before the bot is
created. If validate_config() fails, the bot will not be created.
Add `translate_emoticons` to `prop_types` and `expected_keys`.
Furthermore, create a emoji-translating Markdown inline pattern.
Also use a JavaScript version of `translate_emoticons` and then use
this function during Markdown previews and as a preprocessor. This
is only needed for previews, because usually emoticon translation
happens on the backend after sending.
Add tests for emoticon translation, a settings UI, and a /help/ page
as well.
Tweaked by tabbott to fix various test failurse as well as how this
handles whitespace, requiring emoticons to not have adjacent
characters.
Fixes#1768.
This is necessary for mobile apps to do the right thing when only
RemoteUserBackend is enabled, namely, directly redirect to the
third-party SSO auth site as soon as the user enters the server URL
(no need to display a login form, since it'll be useless).
This got broken at some point when we moved around the context
processing logic for integrations/webhooks. Thankfully, the
context value for external_uri_scheme was only used in a couple of
our less popular integration docs. It should render perfectly now.
Previously, we used to raise an exception if the direct dev login code
path was attempted when:
* we were running under production environment.
* dev. login was not enabled.
Now we redirect to an error page and give an explanatory message to the
user.
Fixes#8249.
This uses an actual query to the backend to check if the subdomain is
available, using the same logic we would use to check when the
subdomain is in fact created.
Shubham Padia
Tim Abbott
Yago González
Yashashvi Dave
Abhilash Verma
Vishnu Ks
Roman Godov
Aditya Bansal
Rohitt Vashishtha
Shubham Dhama
Sarah
Anders Kaseorg
Eeshan Garg
Rishi Gupta
Rhea Parekh
Umair Khan
Raymond Akornor
Steve Howell
Sampriti Panda
Michael J. Sullivan
Angelika Serwa
Vishwesh Jainkuniya
Tarun Kumar
Jack Weatherilt
Harshit Bansal
neiljp (Neil Pilgrim)
Robert Hönig
Marco Burstein
rht
Aastha Gupta