Add JSU Compiler with JMC Backend for C, Python, Java, JS and Perl.

[zx80]

This uses the JSU compiler which uses JMC as a backend for Python, C, Java, JS and Perl.

[jviotti]

Try to merge main back in. The Blaze build should be fine now

[jviotti]

Hey @zx80, just a note that I’ve seen the notifications and will get into it as soon as possible. Quite busy this week with something else + a bit of travel

…

On 12 Mar 2026, at 08:26, Fabien ***@***.***> wrote: @zx80 commented on this pull request. In implementations/jsu-py/version.sh: > @@ -0,0 +1,2 @@ +#! /bin/sh +docker run --rm --entrypoint /app/jsu_version.sh jsonschema-benchmark/jsu-c:latest Python I've done 5 independent dockerfile (I added Java), which as a pain, as I expected. — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

-- Juan Cruz Viotti https://www.sourcemeta.com

[jviotti]

@michaelmior You might want to review this too in the mean-time!

[zx80]

Hmmm, the aggregate tasks fails early on:

Error: An error occurred trying to start process '/usr/bin/bash' with working directory '/home/runner/work/jsonschema-benchmark/jsonschema-benchmark'. Argument list too long

I guess shell limits are reached after adding 5 implementations, some re-engineering will be needed.
After some investigation, it seems that the culprit is:

 printf "${OUTPUTS}" | awk 'NR==1 || !/^implementation,/' > dist/report.csv

The OUTPUTS environment variable contains the whole CSV file contents, which must be over 128 kB…
This should probably be turned into artifacts instead.

[zx80]

I succeeded in updating the CI to use artifact uploads instead of output transfers so as to avoid failing on the maximum size of an environment variable.

Now the aggregate tasks fails on missing crendentials when uploading the results, which is the usual.

[jviotti]

I left a comment regarding pipes but otherwise looks good. @michaelmior can you also do a pass on this, as you wrote the vast majority of the surrounding scripts for measuring, etc?

[zx80]

Hello Juan,

I'm unsure about what I can do to help this PR move forward.

FYI, in complement to JSU Python backend variant and thanks to Julian, the JSU C, Java, JS and Perl backend variants have been integrated (C commit and Java/JS/Perl commit) to bowtie, thus should be listed on the next report update.

The August PR (a resubmission of this May PR that I wrongly thought had been merged) about listing these tools on the JSON Schema website seems as stuck as ever.

[michaelmior]

Overall I think this is looking good although there are several minor things that should be addressed.

[zx80]

Overall I think this is looking good although there are several minor things that should be addressed.

I think that I've addressed everything in the above push.

[jviotti]

I think it's ready!

[zx80]

I think it's ready!

Great! Thanks! When do you think it may be merged?

[jviotti]

@zx80 Its pending @michaelmior 's re-review. I can't merge given the requested changes.

[zx80]

Ok!

[jviotti]

Ping @michaelmior

Changes addressed

[michaelmior]

Sorry for the delay. I resolved my review so feel free to merge! @jviotti note that as an admin, you should be able to dismiss my review yourself to meet the merge requirements.

[jviotti]

@michaelmior Ah, looks like it was not the case, but maybe some repo settings we could tweak. As per my screenshot above, I could not merge indeed.

[michaelmior]

@jviotti Yes, I understand you can't directly bypass merging, but I think maybe you could have dismissed my review which would then have allowed you to merge. Anyway, it's done now :) Thanks @zx80!

[zx80]

Thanks for the reviews and the merge!