Because
- Json schemas were incorrectly failing validation on `firefoxLabsTitle`
and `firefoxLabsDescription` fields. We need to allow them to be `null`
(omitted).
This commit
- Updates `DesktopNimbusExperiment` and `DesktopNimbusExperiment`
classes' fields mentioned above.
- Updates schema version.
- Generates new TS schema.
- Adds new test (fixture) for this use case.
Fixes#11749
Because:
- Experimenter uses a different placeholder feature ID than the
specified in the schemas; and
- using the new split schema in Firefox Desktop would break the world
This commit:
- splits the Firefox Desktop schema into two versions: a strict,
backwards-compatible version (one that enforces the branches[].feature
field is present), and a more lax client version (that does not require
the branches[].feature field);
- updates the strict schema to use the correct placeholder feature ID;
- updates all the fixtures to use the correct placeholder;
- updates all the tests to test against both the strict and lax schemas;
and
- updates the schemas package to 2024.11.4.
Fixes#11717
Because:
- our current pan-application schema currently allows SDK experiments to
validate on Firefox Desktop (because there is no conditional validation
enforcing a relationship between the application field and the branches
field);
- we no longer publish single feature experiments; and
- we no longer support applications that do not support multi-feature
experiments.
This commit:
- splits our NimbusExperiment schema into a DesktopNimbusExperiment and
an SdkNimbusExperiment;
- removes support for mono-feature experiments from both Desktop and SDK
experiment schemas;
- removes some old, mono-feature fixtures that do not validate with our
new schemas; and
- bumps the schemas package version to 2024.11.3 to publish these
changes.
Fixes#11704
Because
- We found a bug where the `firefoxLabsTitle` field was set as required
on the `ExperimentMultiFeatureDesktopBranch` class.
This commit
- Updates the `firefoxLabsTitle` field on the above class to use
`default=None`.
- Updates schema version to `2024.11.2`.
- Updates TS schema
Fixes#11701
Because:
- we previously were using the generated JSONSchema definition of the
localizations field to validate it in Experimenter; and
- we now have pydantic v2 in Experimenter
This commit:
- pulls out the localizations field into a separate model so that we can
use that for valdiation directly in Experimenter; and
- bumps the version to 2024.11.1
Fixes#11673
Because
- We need to update the Nimbus Experiment schema with Firefox Labs
relevant fields.
This commit
- Updates `NimbusExperiment` class in the `experiments.py` module with
these optional fields `isFirefoxLabsOptIn`, `firefoxLabsTitle`,
`firefoxLabsDescription`.
- Updates `ExperimentMultiFeatureDesktopBranch` class in the
`experiments.py` module with the optional field `firefoxLabsTitle`.
- Adds relevant generated json and TS schemas.
- Updates and adds unit tests for `NimbusExperiment`.
- Adds fixtures(json) for the above tests.
Fixes #[11562](https://github.com/mozilla/experimenter/issues/11562)
Because:
- we need the .schema.json files in Desktop;
- they are not currently packaged into either the NPM or PyPI packages;
and
- the generated schemas do not include the custom validation logic
expressed by Pydantic validators
This commit:
- adds JSON schema validation via Pydantic's json_schema_extra hook to
give the JSON schemas validation parity with the Pydantic models and
existing JSON schemas in Desktop;
- generates the JSON schemas and commits them (in schemas/schemas);
- packages the produced schemas into the NPM package;
- packages the produced schemas into the PyPI package (but uncommitted,
since they are duplicates);
- revises the Makefile build steps for schema package generation; and
- bumps the schema package version.
Fixes#11518
Because
- last publish of schemas package to PyPI failed
- failure appears to be related to a bug documented in twine that was
fixed in the latest version
This commit
- upgrades to the latest 5.1.1 version
- bumps version of schemas so it will be published
Fixes#11331
Because
- the schemas package is used by a lot of data tooling
- the data tooling is starting to complain about schemas requiring
pydantic v1 due to other dependencies
- the pydantic-to-typescript package is outdated and does not support
pydantic v2
This commit
- updates the schemas package to v2
- changes the typescript generation to a script internal to experimenter
Fixes#11214
Because
- we added p_value to Statistic but never had time to follow up with
using it
- Jetstream would need to update to add this field to the results schema
- we don't foresee being able to do anything with this anytime soon
- it's easy to add back if we ever get around to using it
This commit
- remove p_value from Statistic
Fixes#11287
Because
- analysis tooling is using python 3.10
- schemas package is using 3.11
- analysis tooling relies on schemas package
This commit
- downgrades schemas package to bring them in sync on python 3.10
Fixes#11266
Because
- randomization unit is configurable
- new profile mgmt feature requires jetstream to behave differently
depending on the randomization unit
- randomization unit options are defined in Experimenter where Jetstream
can't use them
- we don't want to redefined the options in multiple places
This commit
- adds randomization unit enum to schemas package so it can be used by
both Experimenter and Jetstream
Fixes#11244
Because
- we added "all" as a possible user type in precomputed pop sizing data
This commit
- adds "all" as an option to the SizingUserType enum
Fixes#10929
Because
- we want to export metric and image version metadata from jetstream
(https://github.com/mozilla/jetstream/issues/1786)
- schemas_build make command was confusing
This commit
- adds config version to the jetstream metadata schema
- makes schemas_build do the npm and pypi builds
Fixes#10450Fixes#10080
Because
- we want to ensure that Jetstream doesn't publish duplicate results
This commit
- adds validation to the schema to check for duplicate results
- bumps schemas package to version `2024.1.3`
Fixes#10043
Adds field to statistics schema
Because
- It would be useful to data scientists if Jetstream results contained
p-values
This commit
- Adds p-values to the shared schemas
Fixes #github_issue_number
Because
- Some of these enums could be useful in the UI as typescript types
This commit
- Copies the enums defined in Experimenter's Jetstream results ingestion
client to the schemas package
Fixes#10001
Because
* As we add a wider variety of elements to the Nimbus project including
binary packages it becomes increasingly important to have a consistent
environment for each of our modules
* Using Docker allows us to control the environment each module runs in
and use a consistent base layer
* Dockerizing the schemas package will allow us to use identical Python,
Node, and Rust environments as the other modules in Nimbus
This commit
* Wraps the schemas module in a Dockerfile that is consistent with the
other modules in Nimbus
Because:
- the feature schema was causing superfluous errors that weren't really
errors (e.g., complaining that has_exposure must be true/false) when
running into real errors; and
- the hasExposure field is required by desktop;
This commit:
- adds a new Feature type which is a discriminated union, which will
prevent these errors; and
- marks the hasExposure field as a required field.
Because
- the feature manifest schema is used in Experimenter and desktop;
- there are bugs in the schema (e.g., the definition of hasExposure);
and
- the field names in the Python definition are not pythonic
This commit
- adds a copy of the schema from experimenter/features to
mozilla_nimbus_schemas;
- updates the field names with aliases to be more pythonic; and
- adds a unit test to validate all vendored manifests.
Because
* We now have the mozilla-nimbus-schemas package for storing all Nimbus
related schemas
* We should add the experiment schema there and update all components to
use it
This commit
* Adds the Nimbus experiment schema to the nimbus schemas package
* Adds all existing test fixtures from shared
Because
- I changed the name of the generated types file to be consistent with
normal TS type filenames (`*.d.ts`)
- I forgot to reference the new filename in the package.json
This commit
- uses the new filename in the package.json
- bumps schemas version so it will be published
Because
- Both auto-sizing and jetstream use the same logs table
- auto-sizing and jetstream logs will be differentiated by a new
`source` column
This commit
- adds the `source` column to the schema
Because
- we want to use the population sizing schemas for graphql and UI
This commit
- sets up generation of typescript schemas from the pydantic schemas
- sets up publishing the typescript schemas to npm
Herraj Luhano
Beth Rennie
Herraj Luhano
Beth Rennie
Mike Williams
Daniel Berry
Jared Lockhart