Support an asterisk in bothify and optimize

[nineinchnick]

Base::bothify() now replaces an asterisk ('*') with either a random number or a random letter.

I also replaced preg_replace_callback in numerify and lexify as it is slower than iterating over a range of chars found by using strpos and strrpos.

Benchmarks are available here: https://gist.github.com/nineinchnick/84895a560910e3ead9cc

The biggest question is should we worry about UTF-8? I guess in a worst case it could replace the wrong byte. There are a few SO questions how to iterate over UTF-8 strings efficiently.

Should the new helper method Base::replaceWildcard be public?

[fzaninotto]

Why don't you directly pass randomDigit() or randomAscii() here? That's save one function call for each placeholder.

[nineinchnick]

How does it save a function call? For randomLetter(), it will be called once anyway and randomDigit() is not called at all, there's an optimization already that gets many digits from a big number.

[fzaninotto]

Hi, thanks for the patch!

Being utf8-safe also explains why the current version may be slower. Although faster, your method drops an important ability (UTF-8), which is a stopper. I think preg_replcae_callback() with a /u is a must.

[nineinchnick]

I've been thinking about it. We can still use strpos to extract a shorter string to pass to preg_replace_callback(). It also allows to skip the call completely in bothify if there are no '#' or '?'.

I'll benchmark that version and the one with custom utf-8 string iteration.

[nineinchnick]

Btw there should be an unit test to see it fail for special utf-8 strings. Maybe I'll try it with minimaxir/big-list-of-naughty-strings.

[nineinchnick]

There's an answer on SO that says it should be fine to search for an ascii char in an utf-8 string.

[fzaninotto]

Then you'll need a test to prove it :)

[nineinchnick]

Done and it seems to work. Done with https://gist.github.com/nineinchnick/917e644df42ccd62db5c
for #, * and ? on every possible utf-8 char and those files:

• minimaxir/big-list-of-naughty-strings
• bits/UTF-8-Unicode-Test-Documents

[fzaninotto]

I mean a unit test un Faker.

[nineinchnick]

Done.

[fzaninotto]

I meant testing Faker, not PHP. You have to make sure numerify, lexify, and bothify work correctly with UTF-8 strings.

[nineinchnick]

I'm not sure how such a test would look like. Should I run it on that big utf-8 string, replace the single char back to the placeholder and compare?

[fzaninotto]

No, just use a sample UTF-8 string with a couple special chars AND placeholders, and check that the result matches the expected with a regular expression (using /u)

[nineinchnick]

Done. Note that it doesn't matter if you use /u or not.

[fzaninotto]

Great, thanks!