Homec4science
Diffusion libphutil (production) Browse

libphutil/src/parser/PhutilDocblockParser.php
f8af794b2135production

Last Change
Raw File

PhutilDocblockParser.php
View Options

<?php
/**
* Parse a docblock comment from source code into raw text documentation and
* metadata (like "@author" and "@return").
*/
final class PhutilDocblockParser extends Phobject {
public function extractDocblocks($text) {
$blocks = array();
$matches = null;
$match = preg_match_all(
'@(/\*\*.*?\*/)@s',
$text,
$matches,
PREG_OFFSET_CAPTURE | PREG_PATTERN_ORDER);
if (!$match) {
return $blocks;
}
// Build a map of character offset -> line number.
$map = array();
$lines = explode("\n", $text);
$num = 1;
foreach ($lines as $line) {
$len = strlen($line) + 1;
for ($jj = 0; $jj < $len; $jj++) {
$map[] = $num;
}
++$num;
}
foreach ($matches[0] as $match) {
list($data, $offset) = $match;
$blocks[] = array($data, $map[$offset]);
}
return $blocks;
}
public function parse($docblock) {
// Strip off comments.
$docblock = trim($docblock);
$docblock = preg_replace('@^/\*\*@', '', $docblock);
$docblock = preg_replace('@\*/$@', '', $docblock);
$docblock = preg_replace('@^\s*\*@m', '', $docblock);
// Normalize multi-line @specials.
$lines = explode("\n", $docblock);
$last = false;
foreach ($lines as $k => $line) {
// NOTE: We allow "@specials" to be preceded by up to two whitespace
// characters; more than that and we assume the block is a code block.
// Broadly, there's ambiguity between a special like:
//
// <... lots of indentation ...> @author alincoln
//
// ...and a code block like:
//
// <... lots of indentation ...> @def square(x, y):
//
// Because standard practice is to indent the entire block one level,
// we allow that and one additional space before assuming something is
// a code block.
if (preg_match('/^\s{0,2}@\w/i', $line)) {
$last = $k;
$lines[$last] = trim($line);
} else if (preg_match('/^\s*$/', $line)) {
$last = false;
} else if ($last !== false) {
$lines[$last] = $lines[$last].' '.trim($line);
unset($lines[$k]);
}
}
$docblock = implode("\n", $lines);
$special = array();
// Parse @specials.
$matches = null;
$have_specials = preg_match_all(
'/^@([\w-]+)[ \t]*([^\n]*)/m',
$docblock,
$matches,
PREG_SET_ORDER);
if ($have_specials) {
$docblock = preg_replace(
'/^@([\w-]+)[ \t]*([^\n]*)?\n*/m',
'',
$docblock);
foreach ($matches as $match) {
list($_, $type, $data) = $match;
$data = trim($data);
// For flags like "@stable" which don't have any string data, set the
// value to true.
if (!strlen($data)) {
$data = true;
}
if (!isset($special[$type])) {
$special[$type] = $data;
} else {
if (!is_array($special[$type])) {
$special[$type] = (array)$special[$type];
}
$special[$type][] = $data;
}
}
}
// Convert `array(true, true, true)` to `true`.
foreach ($special as $type => $data) {
if (is_array($data)) {
$all_trues = true;
foreach ($data as $value) {
if ($value !== true) {
$all_trues = false;
break;
}
}
if ($all_trues) {
$special[$type] = true;
}
}
}
$docblock = str_replace("\t", ' ', $docblock);
// Smush the whole docblock to the left edge.
$min_indent = 80;
$indent = 0;
foreach (array_filter(explode("\n", $docblock)) as $line) {
for ($ii = 0; $ii < strlen($line); $ii++) {
if ($line[$ii] != ' ') {
break;
}
$indent++;
}
$min_indent = min($indent, $min_indent);
}
$docblock = preg_replace(
'/^'.str_repeat(' ', $min_indent).'/m',
'',
$docblock);
$docblock = rtrim($docblock);
// Trim any empty lines off the front, but leave the indent level if there
// is one.
$docblock = preg_replace('/^\s*\n/', '', $docblock);
return array($docblock, $special);
}
}
c4science · Help