Bug in Underscore Namingstrategy

[boesing]

Hey guys,

in v2, the failing unit test worked as expected.
Since v3, the CamelCaseToUnderscoreFilter adds various underscores around the numeric values within a string.

Any suggestion how this can be fixed without breaking the other tests?

And why was that behavior changed that much?
CamelCaseToSeparator of zend-filter works a lot different than the new logic.

I dont even get the new logic:

-'bfd7b82e9cfceaa82704d1c1_foo'
+'bfd_7_b82e_9_cfceaa_82704_d1c_1_foo'

So why the string is being separated with _ after bfd and not after bfd7b again? x_O

[weierophinney]

I dont even get the new logic:

-'bfd7b82e9cfceaa82704d1c1_foo'
+'bfd_7_b82e_9_cfceaa_82704_d1c_1_foo'

So why the string is being separated with _ after bfd and not after bfd7b again? x_O

It looks like it's separating at numeric boundaries... but, honestly, if that were 100% the case, I'd expect bfd_7_b_82_e_9_cfceaa_82704_d_1_c_1_foo.

What happened for v3 was that we brought the filters into the package, to remove the dependency on zend-filter. It looks like there are differences between zend-filter and what we do here, though, so I'll review the two side-by-side to figure out how and why, and fix it.

Thanks for the failing test!

[boesing]

It looks like it's separating at numeric boundaries... but, honestly, if that were 100% the case, I'd expect bfd_7_b_82_e_9_cfceaa_82704_d_1_c_1_foo.

Thats exactly what I thought, yup.

[weierophinney]

@boesing Can you enable maintainer push rights on your fork? I have a patch I've built against your branch that contains a fix, and I'd like to push back to your branch.

Thanks!

[boesing]

@weierophinney edits from maintainers are enabled in this PR or do I have to do anything else?

[michalbundyra]

@boesing On the side in your PR just check the checkbox:

[boesing]

@boesing On the side in your PR just check the checkbox:

Thats already activated.

[weierophinney]

Really strange - it won't let me push, citing access rights.

I'll push to my own branch and open a new PR, closing this one.

[weierophinney]

#103 builds on your tests; thanks, @boesing !