Comments on empty blocks are dropped.

[tsufeki]

$ php-parse --var-dump '<?php /* a */{}'
====> Code <?php /* a */{}
==> var_dump():
/home/.../bin/php-parse:86:
array(0) {
}

Non-empty blocks work fine. I'm using master.

[tsufeki]

It seems that comments "in the middle" of compound statements are ignored too, for example

<?php if /* a */ (7) {}

[tsufeki]

One more thing: comments are often repeated, i.e. same comment appears in many nodes. This can be seen in tests, for example parser/blockComments.test. Is this intended? If so, what are the rules for repetition? Thanks for your work.

[7fe]

@tsufeki in my opinion <?php if /* a */ (7) {} is acted as intended. I believe the comments are intended for typical documentation which wouldn't include such.

[nikic]

There are three separate issues here:

• Preservation of comments on empty blocks. This has been fixed by d418bf3.
• Duplicate comments. This is a known bug tracked in Comments are assigned multiple times #253.
• Comments in the middle of nodes. These are indeed not represented. As usual, it's still possible to extract them using token offsets and a bit of extra work. Assuming token offsets are enabled in the lexer, something along the lines of (not tested):

// $tokens from Lexer->getTokens()
function getInteriorComments(Node $node, array $tokens) {
    $comments = [];
    $pos = $node->getStartTokenPos();
    foreach ($node->getSubNodeNames() as $name) {
        $subNode = $node->$name;
        if ($subNode instanceof Node) {
            $endPos = backwardsAdjust($tokens, $subNode->getStartTokenPos());
            $comments = array_merge($comments, extractComments($tokens, $pos, $endPos));
            $pos = $subNode->getEndTokenPos() + 1;
        }
    }
    $endPos = $node->getEndTokenPos() + 1;
    $comments = array_merge($comments, extractComments($tokens, $pos, $endPos));
    return $comments;
}

function extractComments(array $tokens, int $startPos, int $endPos) {
    $comments = [];
    for ($pos = $startPos; $pos < $endPos; $pos++) {
        $token = $tokens[$pos];
        if ($token[0] === T_COMMENT) {
            $comments[] = new Comment($token[1]);
        } else if ($token[0] === T_DOC_COMMENT) {
            $comments[] = new Comment\Doc($token[1]);
        }
    }
    return $comments;
}

function backwardsAdjust(array $tokens, int $pos) {
    for (; $pos > 0; $pos--) {
        $token = $tokens[$pos-1];
        if (!in_array($token[0], [T_COMMENT, T_DOC_COMMENT, T_WHITESPACE])) {
            break;
        }
    }
    return $pos;
}

What this code does is to fetch all the comments that are inside the node but not inside a subnode. An additional complication (the backwardsAdjust function) is that we also want to skip the comments that are directly preceding the subnodes, as there will already be associated with the subnodes. (This would have been somewhat easier if comments also stored their starting token offset -- they currently only have the file offset.)