Option to group errors by error type - proposal

[LeonardoCA]

When starting to use PHPStan on an existing project there might be thousands of errors.

In this case I find useful to group the errors by error type to be able to handle similar errors together and to have an overview which errors need attention first.

The challenge is there is no errorType property in AnalysisResult

I wrote a custom formatter, which uses regular expressions on the error messages as a workaround, but it is nasty and not reliable solution

The desired output

List of errors grouped by error type

A compact summary of error types and counts as an overview

I did get this compact report by commenting out one line in my custom formatter. This would need to be a second formater or there needs to be a possibility to pass a parameter to the formater from the command line.

[ondrejmirtes]

See: #1608 (comment)

[DaveLiddament]

Keen to assist with this in anyway I can. (It'd be a very useful feature for a static analysis base liner tool I'm writing.)

As a first attempt at solving this would something as simple as passing in the Rule (when known) to the Error object be satisfactory. (DaveLiddament@1724bbb)

The formatter can then output a string representation of the rule e.g. DaveLiddament@5473e88

If the purpose of the type is for classification rather than showing to the end user then the Rule's FQCN is probably sufficient:

• Using the FQCN guarantees that the string representation is unique.
• No new property (e.g. name) on the Rule object is needed, so it will not be a BC break.

[Wosi]

What is the current state of this feature request? Why didn't @DaveLiddament's changes made it into master? Is there a chance to see grouped errors in the future?

[ondrejmirtes]

Rule class isn't sufficient rule identifier.

I'd like each reported error to be able to tell what identifier to use. Rule class names are implementation details.

It could be done by adding identifier method to RuleErrorBuilder, similar how this was done: #1863

Rules could then opt into providing a custom identifier. Errors reported without an identifier would be grouped as "other".

This feature would also allow to turn off specific reports after the analysis in the config file, or using something else than regexps to ignore errors.

[ThomasLandauer]

I updated the gist by @LeonardoCA (see above) for current versions of phpstan.

It would be nice if something like this could be integrated into core. Use case: If you have many errors, it's easier to fix the exact same issue in 50 files, than opening one file and fixing its 20 different issues.

[ondrejmirtes]

This is mostly a question of setting identifier for all PHPStan-reported errors. There are already some errors marked by that, see where RuleErrorBuilder::identifier() is called.