([bc]+ { return true; }) rule is not parsed about pegjs

[shamansir]

Additional Info:

In my fork this code also fails:

var actionKnowsEndPositionInsideParser = PEG.buildParser(
    'start = "e" "d" ([bc]* { return 4; }) "a"'
);
parses(actionKnowsEndPositionInsideParser, "edcba", ["e", "d", 4, "a"]);

And I can help it to pass with changing [bc]* to [b-c]*.

However, both this variants ([bc]* and [b-c]*) pass ok in a clean clone of https://github.com/dmajda/pegjs.git

[shamansir]

I've tried to go deep with these issues (issue #55 is directly related), and, it seems, this is the very deep thing. The parser result block in my case is:

        if (/^[bc]/.test(input.charAt(pos))) {
                result3 = input.charAt(pos);
          pos++;
        } else {
          result3 = null;
          if (reportFailures === 0) {
            matchFailed("[bc]");
          }
        }
          }
        }
        while (result3 !== null) {
          result2.push(result3);
          if (/^[bc]/.test(input.charAt(pos))) {
            result3 = input.charAt(pos);
            pos++;
          } else {
            result3 = null;
            if (reportFailures === 0) {
              matchFailed("[bc]");
            }
          }
        }
        _chunk.pos = pos1;
              _chunk.end = pos;
        _chunk.match = input.substring(pos1,pos);
        i  f (result2 !== null) {
          result2 = (function(_chunk) { return _chunk.end; })(_chunk);
        }

You see the incorrect indent at start and broken i f in the end.

• May be the unindent/indent mechanics fail in Codie, but why they fail only for this concrete rule among all 480+?
• May be the problem is in literal (more possible) or zero_or_more (less possible, one_or_more fails in your branch) templates, but the code for them seems right (I've tried grouping if-s and aligning indents in different ways and nothing helped)
• May be the problem is somewhere else, only you know where)

In conclusion, I suppose that the reason is that complex if-else-nested rule in literal template (it was introduced only in last commits with ignoreCase flag) and indentation fails on parsing of that.

[dmajda]

This is a duplicate of #55 which was fixed. Thanks for digging into the issue.