Comparing changes

…is set, also in case of missing input exceptions (#3786) ### Description  ### QC  * [x] The PR contains a test case for the changes or the changes are already covered by an existing test case. * [x] The documentation (`docs/`) is updated to reflect the changes or this is not necessary (e.g. if the change does neither modify the language nor the behavior or functionalities of Snakemake).  ## Summary by CodeRabbit * **New Features** * Formatted file references now clearly indicate when a file is in storage by appending “in storage” where applicable. * **Bug Fixes** * Standardized the return of formatted file paths: in non-input/output contexts, formatting now consistently returns the annotated string representation, ensuring predictable display and messaging across the app.

This PR fixes a regression introduced in Snakemake 9.6.0 where relative `$ref` references in local config schemas were not correctly resolved during validation. This regression was introduced when refactoring the config validation from the (deprecated) `RefResolver` to the (current) `resolving` API in #3420. The two required changes were: 1. Injection of an absolute `file://` `$id` in schemas to enforce proper local reference resolution. 2. A follow-up update to retrieve_uri to correctly load subschemas (as it now receives a `file://` URI). In addition, to prevent future regressions, this PR adds tests for (nested) `$ref`s, `$ref` with remote `$id` (see below), fragment references, `allOf`/`anyOf` defaults, and `$def`s. This commit restores pre-9.6.0 behavior and prevents ValidationError when nested schemas are referenced. Caveat: This PR intentionally does *not* restore one specific aspect of the old (`RefResolver` based) implementation: If a schema file contains an explicit `$id` pointing to a remote copy of the schema, the old implementation would fetch any references (in- or external) from that remote URI. Instead, the implementation introduced by this PR always resolves references based on the local schema file. This behaviour is demonstrated (and validated) with the remote `$id` test mentioned above. The following checks have been performed to ensure the correctness of this patch: 1. All tests (but the remote `$id` one, see above) were successfully run at commit [`960f6a89eaa31da6014e810dfcf08f635ac03a6e`](960f6a8) (last ancestor of current `main` that still uses the old `RefResolver` implementation), proving that they cover pre-9.6.0 behaviour. 2. Re-running the same tests at commit [`cf724272eefcd9b30fb625d2e16380727bef9c3e`](cf72427) (direct child of the above, introducing the changes from PR #3420), the three (nested) `$ref`s test fail, verifying this commit/PR caused the regression and the tests added in this commit catch that regression. 3. The same tests still fail at commit [`2eb9079dae29f271aeacff75e59f6c2f6a0d352d`](2eb9079) (current `main`), demonstrating that the regression has not been fixed yet. 4. All tests pass after applying the changes from this PR, validating it fixes the regression. ### QC * [x] The PR contains a test case for the changes or the changes are already covered by an existing test case. * [x] The documentation (`docs/`) is updated to reflect the changes or this is not necessary (e.g. if the change does neither modify the language nor the behavior or functionalities of Snakemake). #### Notes: This is my first contribution to the project. Thus, I am not familiar with the codebase. It is also my first time dealing with `jsonschema` and the both, its (deprecated) `RefResolver` API (to analyse the old implementation and design tests accordingly), and its current `resolving` API (to fix the regression). Please let me know if you have any adjustments you want/need me to do. Also, I started from a standalone POC script to make sure I understand the API, defined tests and then worked on the actual fix off that. Initially, I overestimated the power of pre-9.6.0's implementation of the `validate` function and designed a test that tested defaults propagation across (nested) `$ref`s. Thus, my intermediate solution was much more complex (and powerful) than this simple (minimal) fix. I do have a local branch with that work in case anybody is interested in implementing more powerful defaults value handling during config validation. Please reach out if there is interest. --- fixes #3648  ## Summary by CodeRabbit * **Bug Fixes** * Schema resolution now enforces local resolution of relative $ref references, improving validation and default assignment across nested refs, fragments, allOf/anyOf/defs; may change behavior for schemas that previously relied on remote resolution. * **Tests** * Added comprehensive tests and fixtures covering relative/fragment references, nested schemas, allOf/anyOf/defs, default handling, and error scenarios.

🤖 I have created a release *beep* *boop* --- ## [9.13.1](v9.13.0...v9.13.1) (2025-10-11) ### Bug Fixes * **config:** restore $ref resolution for schemas ([#3777](#3777)) ([a3e0939](a3e0939)) * fix iofile formatting to always display the storage query if it is set, also in case of missing input exceptions ([#3786](#3786)) ([b1c29fa](b1c29fa)) --- This PR was generated with [Release Please](https://github.com/googleapis/release-please). See [documentation](https://github.com/googleapis/release-please#release-please).

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Comparing changes

Open a pull request

Commits on Oct 11, 2025

This comparison is taking too long to generate.

Uh oh!