PHP - preg_replace_callback for camelCasing

Question

I have the following content

"aa_bb" : "foo"
"pp_Qq" : "bar"
"Xx_yY_zz" : "foobar"

And I want to convert the content on the left side to camelCase

"aaBb" : "foo"
"ppQq" : "bar"
"xxYyZz" : "foobar"

And the code:

// selects the left part
$newString = preg_replace_callback("/\"(.*?)\"(.*?):/", function($matches) {        
    // selects the characters following underscores
    $matches[1] = preg_replace_callback("/_(.?)/", function($matches) {
        //removes the underscore and uppercases the character
        return strtoupper($matches[1]);
    }, $matches[1]);

    // lowercases the first character before returning
    return "\"".lcfirst($matches[1])."\" : ".$matches[2];
}, $string);

Can this code be simplified?

Note: The content will always be a single string.

Answer 1

First, since you already have a working code you want to improve, consider to post your question in code review instead of stackoverflow next time.

Let's start to improve your original approach:

$result = preg_replace_callback('~"[^"]*"\s*:~', function ($m) {
    return preg_replace_callback('~_+(.?)~', function ($n) {
        return strtoupper($n[1]);
    }, strtolower($m[0]));
}, $str);

pro: patterns are relatively simple and the idea is easy to understand.
cons: nested preg_replace_callback's may hurt the eyes.

After this eyes warm-up exercice, we can try a \G based pattern approach:

$pattern = '~(?|\G(?!^)_([^_"]*)|("(?=[^"]*"\s*:)[^_"]*))~';
$result = preg_replace_callback($pattern, function ($m) {
    return ucfirst(strtolower($m[1]));
}, $str);

pro: the code is shorter, no need to use two preg_replace_callback's.
cons: the pattern is from far more complicated.

notice: When you write a long pattern, nothing forbids to use the free-spacing mode with the x modifier and to put comments:

$pattern = '~
(?| # branch reset group: in which capture groups have the same number
    \G # contigous to the last successful match
    (?!^) # but not at the start of the string    
    _
    ( [^_"]* ) # capture group 1
  |
    ( # capture group 1
        "
        (?=[^"]*"\s*:) # lookahead to check if it is the "key part"
        [^_"]*
    )
)
~x';

Is there compromises between these two extremes, and what is the good one? Two suggestions:

$result = preg_replace_callback('~"[^"]+"\s*:~', function ($m) {
    return array_reduce(explode('_', strtolower($m[0])), function ($c, $i) {
        return $c . ucfirst($i);
    });
}, $str);

pro: minimal use of regex.
cons: needs two callback functions except that this time the second one is called by array_reduce and not by preg_replace_callback.

$result = preg_replace_callback('~["_][^"_]*(?=[^"]*"\s*:)~', function ($m) {
    return ucfirst(strtolower(ltrim($m[0], '_')));
}, $str);

pro: the pattern is relatively simple and the callback function stays simple too. It looks like a good compromise.
cons: the pattern isn't very constrictive (but should suffice for your use case)

pattern description: the pattern looks for a _ or a " and matches following characters that aren't a _ or a ". A lookahead assertion then checks that these characters are inside the key part looking for a closing quote and colon. The match result is always like _aBc or "aBc (underscores are trimmed on the left in the callback function and " stays the same after applying ucfirst).

pattern details:

["_] # one " or _
[^"_]* # zero or more characters that aren't " or _
(?= # open a lookahead assertion (followed with)
    [^"]* # all that isn't a "
    " # a literal "
    \s* # eventual whitespaces
    : # a literal :
) # close the lookahead assertion

There's no good answer and what looks simple or complicated really depends on the reader.

Answer 2

You might make use of preg_replace_callback in combination with the \G anchor and capturing groups.

(?:"\K([^_\r\n]+)|\G(?!^))(?=[^":\r\n]*")(?=[^:\r\n]*:)_?([a-zA-Z])([^"_\r\n]*)

In parts

• (?: Non capturing group
  • "\K([^_\r\n]+) Match ", capture group 1 match 1+ times any char except _ or newline
  • | Or
  • \G(?!^) Assert position at the previous match, not at the start
• ) Close group
• (?=[^":\r\n]*") Positive lookahead, assert "
• (?=[^:\r\n]*:) Positive lookahead, assert :
• _? Match optional _
• ([a-zA-Z]) Capture group 2 match a-zA-Z
• ([^"_\r\n]*) Capture group 3 match 0+ times any char except _ or newline

In the replacement concatenate a combination of strtolower and strtoupper using the 3 capturing groups.

Regex demo

For example

$re = '/(?:"\K([^_\r\n]+)|\G(?!^))(?=[^":\r\n]*")(?=[^:\r\n]*:)_?([a-zA-Z])([^"_\r\n]*)/';
$str = '"aa_bb" : "foo"

"pp_Qq" : "bar"

"Xx_yY_zz" : "foobar"
"Xx_yYyyyyyYyY_zz_a" : "foobar"';

$result =  preg_replace_callback($re, function($matches) {
    return strtolower($matches[1]) . strtoupper($matches[2]) . strtolower($matches[3]);
}, $str);

echo $result;

Output

"aaBb" : "foo"

"ppQq" : "bar"

"xxYyZz" : "foobar"
"xxYyyyyyyyyyZzA" : "foobar"

Php demo