Precompiled 2024-05-02-3f5dcf1

Search deleted terms tracker (#25254)

This PR adds the logic for the search deleted terms tracker. We'll write three tracker files - id tracker, deleted bitset, and deleted terms since not all of them are needed everywhere. The deleted terms file format is pretty similar to id tracker but instead of using perfect hashing, we store the term ordinals and number of deleted documents corresponding to that term in separate structures: `EliasFano` for the `TermOrdinal`s because they are monotonic increasing integers, and `DacsOpt` for the counts of deleted documents because they will be small positive integers.

GitOrigin-RevId: 8e42c994af566b27d0a16219898e36459ec19af1

Precompiled 2024-05-01-6282906

Propagate action cancellation for Funrun (#24942)

This would stop the actions if backend goes away but otherwise result in the same behavior as today.

GitOrigin-RevId: 9ab50448ef5afde7815e2cdc9155dad46aedd117

Precompiled 2024-04-30-2e7f0eb

[modules] populate cache with multiple values (#25152)

when we start storing modules in S3, we will want to populate the entire module cache at once by reading the S3 zip file once. To do this, we should add a method to AsyncLru `get_and_prepopulate` that has the same functionality as `get` but the ValueGenerator can return values for additional keys, prepopulating them for future gets.

Started using this method in `ModuleCacheWorker` with no benefits over the existing API -- it's just exercising the codepath in preparation for the S3 use-case.
Also add a unit test.

GitOrigin-RevId: 1191f1558778384cee7690e3f4fb9fcc17ffb07a

Precompiled 2024-04-29-5b5828c

Ignore convex/#.* files which are likely emacs tempfiles (#25144)

GitOrigin-RevId: fb0d04a7776731303c11eb1ecdb5c74d0b20f7dc

Precompiled 2024-04-26-560e5a3

Clean shutdown the event logger for usage tracking (#24947)

GitOrigin-RevId: a86742f2a39d415bf12d08c35cafa093e14308ac

Precompiled 2024-04-25-03a1142

remove `_modules.deleted` field (#25043)

we have removed all documents for which `deleted` is true. clean up the field

GitOrigin-RevId: 7195e013678db9627488482e108008b39fadf8cd

Precompiled 2024-04-24-c9ec833

Bump rustls (#24948)

GitOrigin-RevId: 2fbf571f6638426109081e1f2d3bd826ee887009

Precompiled 2024-04-23-788afe4

Bump the pnpm version (minor/patch only) (#24904)

GitOrigin-RevId: 5dd133ea33e219f9b8773f06fd243bd9d18d9c84

Precompiled 2024-04-22-0364e93

parallel storage.getUrl (#24821)

allow `ctx.storage.getUrl` in queries and mutations to run in parallel with itself, as a batch.

note this removes a couple of the sanity checks, because i don't think they're necessary and with batching they're harder: (1) removes the assertion that there's at most one stored file with a given storage id, (2) removes the assertion that `entry.storage_id == storage_id` when `entry` was fetched with an equality index filter on `storage_id`

i wanted to make it also run in parallel with `db.get` and `db.query` but that turns out to be tricky because those need to use `DeveloperQuery` while file storage urls needs to use `ResolvedQuery`.

GitOrigin-RevId: 41e686efe0b36c60d9e91a99a5cfc8bc4d3f56d3

Precompiled 2024-04-19-32883ae

Port even more tests to isolate2 (#24833)

GitOrigin-RevId: f9b8cf1b9fe829748a21f48a9cc36ca4197009dd