Shields: Numeric CPAN versions show as invalid

[dmtucker]

Some projects on CPAN have numeric versions, and they show up as invalid on the CPAN badge.

For example, https://fastapi.metacpan.org/v1/release/App-Netsync:
The license seems fine,
but the version won't show (since fixed).

Here are some more examples of projects that don't use strings:
https://fastapi.metacpan.org/v1/release/Async-Interrupt
https://fastapi.metacpan.org/v1/release/Crypt-SSSS
https://fastapi.metacpan.org/v1/release/Games-Literati
https://fastapi.metacpan.org/v1/release/MojoX-Redis

Here are some examples of projects that use strings:
https://fastapi.metacpan.org/v1/release/Apache-SPARQL
https://fastapi.metacpan.org/v1/release/App-gh
https://fastapi.metacpan.org/v1/release/App-Netdisco
https://fastapi.metacpan.org/v1/release/Test-CheckManifest

I suspect this is because versionColor raises a TypeError when trying to call toLowerCase:
https://github.com/badges/shields/blob/435903ae4a170ffce188141f9c4ddafbc6565a43/lib/color-formatters.js#L14

var version = 3.01
version.toLowerCase()  // TypeError: version.toLowerCase is not a function

Since the MetaCPAN API offers version_numified alongside the version, I would guess this is actually a bug on their end, but the badges could probably work around it pretty easily.

[PyvesB]

Hello @dmtucker !

Thank you for spotting and reporting this. Even though it indeed seems like a bug on the CPAN end, I would be inclined to also cover this in Shields, as this may well happen with other APIs in the future. Does anyone else in the team have an opinion on this matter?

Cheers,

Pyves

[paulmelnikow]

Yea, agreed, let's make the version color formatter accept numbers. I wonder if it this is a regression from the changes we made in 7039e68…

[PyvesB]

@dmtucker would you like to work on a fix in Shields and submit a pull request? I'm happy to tackle it otherwise. 😉

[dmtucker]

@PyvesB Sounds interesting! I'm a bit out of my depth here, though, and I'm not sure when I could get to it. So, please, feel free!

[PyvesB]

I have just merged #1615, it will hopefully be released to production in the coming days/weeks. Here's the result in the staging environment for one of the previously failing badges:

@dmtucker do you think there's anything to add here?

[dmtucker]

Great! Nope, I don't think so. I suspected there was a similar condition in the version text logic, so I'm glad #1620 has gone in too :+1: Thanks for doing those!

[PyvesB]

Awesome, closing the issue then, thanks again for reporting it!

[pryrt]

Sorry to re-vivify this old, closed issue... but since I found it because I was trying to add badges to my Games-Literati (which happened to be one of the broken ones mentioned above!) and found this issue when trying to figure out why the badge doesn't work: is the "invalid response data" the _fix_ from above, or did this bug come back?

• https://fastapi.metacpan.org/v1/release/Games-Literati (broken)
  vs
• https://fastapi.metacpan.org/v1/release/Math-PRBS (working)

Add a screenshot, since the live links may change over time:

Since I am doing a release soon anyway, I might just go ahead and make my version scheme match my projects that are working, but thought I'd ask

(let me know if I need to create a new issue instead)

[chris48s]

The error being thrown here is ValidationError: "version" must be a string
I think we need to relax the schema in https://github.com/badges/shields/blob/fafb22efee005122be095c4f74212f01df92c139/services/cpan/cpan.js#L7

[PyvesB]

Thanks for the report, I'll look into this again.