Add pre-commit hook config

[AleksaC]

Hi, I think it would be useful to add support for jsonschema as pre-commit hook. I think this would be more convenient than using local hooks or creating a mirror repo.

With the way I set it up, the hook can be used with a config like this:

repos:
  - repo: https://github.com/AleksaC/jsonschema
    rev: 639bc64d82
    hooks:
      - id: jsonschema
        files: |
          (?x)(
            ^valid.json$ |
            ^invalid.json$ |
            ^schema.json$
          )
        args: ["-i", "valid.json", "-i", "invalid.json", "schema.json"]

      - id: jsonschema
        files: |
          (?x)(
            ^valid2.json$ |
            ^invalid2.json$ |
            ^schema2.json$
          )
        args: ["-i", "valid2.json", "-i", "invalid2.json", "schema2.json"]

[codecov]

Codecov Report

Merging #765 (639bc64) into main (4003667) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #765   +/-   ##
=======================================
  Coverage   96.37%   96.37%           
=======================================
  Files          17       17           
  Lines        2703     2703           
  Branches      309      309           
=======================================
  Hits         2605     2605           
  Misses         79       79           
  Partials       19       19           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4003667...639bc64. Read the comment docs.

[Julian]

Hi there. Sounds great, thanks! Will merge and see what feedback we get.

[janosh]

@Julian One piece of feedback (from @asottile, the creator of pre-commit):

this hook is a little bit odd -- it doesn't quite meet the normal invocation structure of pre-commit hooks which accepts [file names as] positional arguments -- I wonder if there's something that can be improved there so the configuration is more easily usable

pre-commit/pre-commit.com#569 (comment)

[Julian]

@janosh Happy to see any improvements -- if I'm honest I've never used jsonschema's own hook myself. If there's a way you see to improve it I'm all for it (otherwise if you don't mind opening a ticket at least to track the feedback I trust Anthony enough to know if something's wrong :)

[janosh]

I'm not actually sure what @asottile is proposing. Presumably, the cleanest CLI interface for pre-commit would be if it accepted the hook file names as positional arguments and the schema file through a flag, i.e.

jsonschema-hook --schema schema.json **/*.json

Not sure, how you would translate between that and jsonschema's current interface though.

[Julian]

I doubt I'll have any time to look at this myself, but that doesn't seem right to me exactly as-is, I doubt the default behavior should be "assume all JSON files in a repo conform to a single schema".

Instead probably someone using the hook should be providing a mapping between schema files and instances to validate under them.

But yeah I'm open to proposals, and yes it seems like a wrapper script meant for pre-commit might be a necessary piece.

[asottile]

I think the interface @janosh suggests is right, but the default configuration should still require the user to set up files and args to filter to the targetted files properly and to select the right schema

something like:

-   id: jsonschema-v2
    name: jsonschema
    entry: jsonschema-hook
    language: python
    # require the user to override args to set `[--schema, somefile.json]`
    # require the user to override `files` to filter to the target files
    files: ^$ 

or ~something like that

in this jsonschema-hook would be a script which has the following "interface"

jsonschema-hook --schema SCHEMA [other options] [file [file [...]]]

[AleksaC]

Hi, as the author of this pull request I'd like to chime in. First of all sorry for not giving more thought to this, it worked well for my use case, but I realize that it isn't ideal. I knew that there wasn't a way to use the current cli with pre-commit in idiomatic way, I just assumed that changing the cli for pre-commit wasn't something you'd want to do.

There is still one thing that isn't addressed - if you change the schema you'd probably want to run the hook on all the files it is supposed to check to make sure there aren't any discrepancies. With the wrapper proposed above it would be solved by defining the hook twice, the second definition being something like this:

hooks:
  - id: jsonschema-v2
    files: ^schema.json$
    args: [--schema schema.json, ...files]
    pass_filenames: false

Anyway if you want to add a wrapper for pre-commit, I'd be willing to do that as it's sort of my mess to clean up 😄

[Julian]

@AleksaC it's never a mess, thanks for the original contribution, it wouldn't have happened without you submitting it in the first place!

And indeed I don't want to change the CLI for pre-commit, but a separate (private) wrapper is possibly fine.

It'd be good to have two pre-commit users agree / propose some behavior for the hook -- perhaps you and @janosh, though certainly @asottile gets a big vote if he feels like guiding us further.

I'm just being realistic when I say I probably won't have brainpower to dedicate at the minute, so I can chime in with stuff like the above but likely not help think through what the thing looks like until I see a PR.

[janosh]

@AleksaC I second @Julian, there's nothing to be blamed for. On the contrary, thanks for getting the ball rolling. Happy to help with the wrapper either through PR or comments.