This makes it much more clear that this feature does JSON encoding,
which previously was only indicated in the documentation.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
These weren’t wrong since orjson.JSONDecodeError subclasses
json.JSONDecodeError which subclasses ValueError, but the more
specific ones express the intention more clearly.
(ujson raised ValueError directly, as did json in Python 2.)
Signed-off-by: Anders Kaseorg <anders@zulip.com>
We only use this in a few places, but they're really important places
for understanding the types in the codebase, and so it's worth having
a bit of expository documentation explaining how we use it.
(And I expect we'll add more with time).
Old: a validator returns None on success and returns an error string
on error.
New: a validator returns the validated value on success and raises
ValidationError on error.
This allows mypy to catch mismatches between the annotated type of a
REQ parameter and the type that the validator actually validates.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Use read-only types (List ↦ Sequence, Dict ↦ Mapping, Set ↦
AbstractSet) to guard against accidental mutation of the default
value.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
Fixes#2665.
Regenerated by tabbott with `lint --fix` after a rebase and change in
parameters.
Note from tabbott: In a few cases, this converts technical debt in the
form of unsorted imports into different technical debt in the form of
our largest files having very long, ugly import sequences at the
start. I expect this change will increase pressure for us to split
those files, which isn't a bad thing.
Signed-off-by: Anders Kaseorg <anders@zulip.com>
‘req_var in request.GET’ was previously believed to be slow from
profiling results. However, the real explanation for those profiling
results is that WSGIRequest.GET is a lazy cached property, so there’s
no reason to avoid it if we’re accessing request.GET anyway.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
A sloppy implementation of the main has_request_variables wrapper
function meant that it did two very inefficient things:
* To combine together the GET and POST parameters, it would make a
copy of the request.GET QueryDict object, which combined with the
fact that these objects are slow to access, consumed about 90us per
argument.
* Doing this in a loop (one time per argument), rather than once,
which resulted in us doing this 11 times for a `GET /events` query.
Fixing this to just make a dictionary and combine things with some
small loops saved about 1 millisecond from the total runtime of GET
/events (for comparison, the total actual work of that view function
is about 700ms).
We need to fix at least one test that used a bad mock HttpRequest
object that didn't have a .GET property.
Then, find and fix a predictable number of previous misuses.
With a small change by tabbott to preserve backwards compatibility for
sending `yes` for the `forged` field.
Signed-off-by: Anders Kaseorg <anders@zulipchat.com>
Most of the failures were due to parameters that are not intended to
be used by third-party code, so the correct fix for those was the set
intentionally_undocumented=True.
Fixes#12969.
Previously, our OpenAPI documentation validation was failing for some
endpoints because it didn't account for the `in: path` type of
parameter, resulting in a mismatch between what was declared via REQ
and what was declared in the OpenAPI docs.
We fix this by excluding the path type parameters in both places from
what's considered by documentation using the `path_only` flag.
I doubt this is the correct long-term fix; in particular, I don't
think we're actually running the validators for these path-only
parameters. The examples that exist today are all IDs with validators
for being non-negative numbers, but longer-term I think we'll want to
do something different (possibly at the REQ layer, see the TODO).
By importing a few view modules in the validation test itself we
can remove a few endpoints which were marked as buggy. What was
happening was that the view functions weren't imported and hence
the arguments map was not filled. Thus the test complained that
there was documentation for request parameters that seemed to be
missing in the code. Also, for the events register endpoint, we
have renamed one of the documented request parameters from
"stream" to "topic" (the API itself was not modified though).
We add a new "documentation_pending" attribute to req variables
so that any arguments not currently documented but should be
documented can be properly accounted for.
This will allow us to mark a REQ variable as intentionally
undocumented. With this, we can remove some of the endpoints marked
as "buggy" even though they're not actually buggy, we just needed to
specify certain parameters as intentionally undocumented (e.g. the
stream_id for the /users/me/subscriptions/muted_topics endpoint.)
Any REQ variable with intentionally_undocumentated set to True
will not be added to the arguments_map data structure.
For some of the other "buggy" endpoints, we would want to mark the
entire endpoint as being undocumented intentionally via. the urls.py
file.
This validation is incomplete, in large part because of the long list
of TODOs in this code. But this test should provide a ton of support
for us in avoiding regressions as we work towards having complete API
documentation.
See https://github.com/zulip/zulip/issues/12521 for a bunch of
follow-up improvements.
Our webhook-errors.log file is riddled with exceptions that are
logged when a webhook is incorrectly configured to send data in
a non-JSON format. To avoid this, api_key_only_webhook_view
now supports an additional argument, notify_bot_owner_on_invalid_json.
This argument, when True, will send a PM notification to the bot's
owner notifying them of the configuration issue.
This is helpful for cases where an argument is supposed to be a normal
string, and we want to use a Zulip validator function to do basic
things like check its length.
This provides the main infrastructure for fixing #5598. From here,
it's a matter of on the one hand upgrading exception handlers -- the
many except-blocks in the codebase that look for JsonableError -- to
look beyond the string `msg` and pass on the machine-readable full
error information to their various downstream recipients, and on the
other hand adjusting places where we raise errors to take advantage
of this mechanism to give the errors structured details.
In an ideal future, I think all exception handlers that look (or
should look) for a JsonableError would use its contents in structured
form, never mentioning `msg`; but the majority of error sites might
continue to just instantiate JsonableError with a string message. The
latter is the simplest thing to do, and probably most error types will
never have code looking for them specifically.
Because the new API refactors the `to_json_error_msg` method which was
designed for subclasses to override, update the 4 subclasses that did
so to take full advantage of the new API instead.
With #5598 there will soon be an application-level error code
optionally associated with a `JsonableError`, so rename this
field to make clear that it specifically refers to an
HTTP status code.
Also take this opportunity to eliminate most of the places
that refer to it, which only do so to repeat the default value.
The file `zerver/lib/request.py` doesn't have type annotations
of its own; if they did, they would duplicate the annotations that
exist in its stub file `zerver/lib/request.pyi`. The latter exists
so that we can provide types for the highly dynamic `REQ` and
`has_request_variables`, which are beyond the type-checker's ken
to type-check, but we should minimize the scope of code that gets
that kind of treatment and `JsonableError` is not at all the sort of
code that needs it.
So move the definition of `JsonableError` into a file that does
get type-checked.
In doing so, the type-checker points out one issue already:
`__str__` should return a `str`, but we had it returning a `Text`,
which on Python 2 is not the same thing. Indeed, because the
message we pass to the `JsonableError` constructor is generally
translated, it may well be a Unicode string stuffed full of
non-ASCII characters. This is potentially a bit of a landmine.
But (a) it can only possibly matter in Python 2 which we intend to
be off before long, and (b) AFAIK it hasn't been biting us in
practice, so we've probably reasonably well worked around it where
it could matter. Leave it as is.
The whole thing is an error, so "message" is a more apt word for the
error message specifically. We abbreviate that as `msg` in the actual
HTTP responses and in the signatures of `json_error` and friends, so
do the same here.
Anders Kaseorg
Alex Vandiver
Tim Abbott
Anders Kaseorg
Tim Abbott
Hemanth V. Alluri
Shoumorup
Eeshan Garg
Steve Howell
neiljp (Neil Pilgrim)
Viraat Chandra
rht
Greg Price
Yago González
Aditya Bansal
sinwar