Page MenuHomePhabricator

Add "phutil_describe_type()" and clean up a few obscure things in "libphutil/"

Authored by epriestley on Feb 26 2019, 3:13 AM.



See PHI1096. With some frequency, I want to raise an exception like "expected method getSomething() to return an object of type X, but it actually returned a thing of type Y".

There's no great builtin function for producing a good value for "Y", the unexpected value's type. "get_type()" is somewhat close, but just returns "object" for objects, when we'd prefer to return the object's class name since this is significantly more useful.

We have a pretty good method for producing "Y" in "libphutil/", but it's hard to remember: PhutilTypeSpec::getTypeOf($value). Wrap it in a convenient name, phutil_describe_type().

Also, provide phutil_is_natural_list() for simplifying the array_keys(...) === range(...) test that we have in a handful of cases.

Finally, clean up a couple of error behaviors that give you a bad/misleading error if you make a syntax error in "utils.php" specifically.

Test Plan

Ran unit tests, see also next diff.

Diff Detail

rPHU libphutil
Automatic diff as part of commit; lint not applicable.
Automatic diff as part of commit; unit tests not applicable.

Event Timeline

epriestley created this revision.Feb 26 2019, 3:13 AM
epriestley requested review of this revision.Feb 26 2019, 3:14 AM
amckinley accepted this revision.Feb 26 2019, 3:34 PM
amckinley added inline comments.

Unfinished thought.

This revision is now accepted and ready to land.Feb 26 2019, 3:34 PM
  • One more similar language tweak.
  • Fix sentence fragment.
This revision was automatically updated to reflect the committed changes.
epriestley added inline comments.Mar 9 2019, 1:58 PM

This accidentally fixed a bug, since PHP is a lovable trickster:

$ php -r 'echo (array("x") == array(0)) ? "SAME ARRAY" : "DIFFERENT";'

That is, this should have been ===.

This bug isn't a huge issue, but it does change the signature algorithm.

This [signature] bug isn't a huge issue

Specifically, I mean it isn't a huge issue since we only use it internally today. If this was in widespread use it could be a real problem, since it would mean the signature algorithm doesn't protect against a MITM attacker very well, but you'd have to be inside the hosted cluster to do anything MITM-flavored today.