Audit functionality

[swissspidy]

Just thinking out loud after reading through https://core.trac.wordpress.org/ticket/44139.

We might be able to generate some warnings when a string occurs multiple times in the source code but with multiple translator comments. That‘s usually an indication that these should be separate strings with different contexts.

[swissspidy]

I'm wondering what the best way would be to get these warnings (and other errors) in a machine-readable manner.

Use case: we currently programmatically extract strings from our projects after each push to GitHub and import them into GlotPress. On success, we send a notification to our Slack channel. Now I want to do the same for errors as well.

Do I need to parse the terminal output using regex, or hold the messages in MakePotCommand and print them at the end in the desired format, or is there a better way?

Many commands have a --format argument, but AFAIK this is only used to print things like objects or lists in a specific format, not error messages.

@schlessera Would love to hear your insights on this :-)

[schlessera]

I've already helped build a similar thing here: https://github.com/fklein-lu/checker-cli/tree/master/src

Basically, it is a generalized pipeline running one or more Checks, and each Check can add issues to a container, like shown here: https://github.com/fklein-lu/checker-cli/blob/master/src/Check/RequiredFilesCheck.php#L34-L42

The Issues contain all the data that is needed for recording problems. Also, they know how to render themselves into text or JSON: https://github.com/fklein-lu/checker-cli/blob/master/src/Issue.php#L76-L101

With this object model, you can easily write multiple ReportGenerators, which could be used for providing different --format=<format> implementations: https://github.com/fklein-lu/checker-cli/tree/master/src/Report

[swissspidy]

Very helpful, thanks!

Will try to explore this concept a bit for the make-pot command so we could eventually add a --format argument to it.

In the mean time I guess we'll resort to parsing the output on our own in our script.