Explicit conflict marker detection

[koppor]

Fixes JabRef#9167

WIP, because the parsing architecture is a bit complicated here.

We cannot "just" read the whole file, because it could be very slow when reading large data bases.

[koppor]

Two implementation ideas:

1. Dive into the grammar and add "exception" paths for conflict markers
2. Do a pre-reading of the file and check for conflict markers
3. Spawn a parallel thread checking for conflict markers. If the normal reading thread returns, with an error, check what the checking thread said.

Think, we could go for 2 even though file loading might be slower?! -- https://www.amitph.com/java-read-write-large-files-efficiently/

[HoussemNasri]

I got the test passing by checking for git markers inside the read method after consuming a newline character.
I also called checkForGitMarkers at the beginning of the file to ensure it's called on the first line.

private int read() throws IOException {
        int character = pushbackReader.read();

        if (!isEOFCharacter(character)) {
            pureTextFromFile.offerLast((char) character);
        }
        if (character == '\n') {
            line++;
            checkForGitConflictMarker();
        }
        return character;
    }

This is the logic. It looks for a line that starts with the 'ours' marker, which is represented by the symbol <<<<<<<. Then it continues to skip lines until it reaches the 'theirs' marker >>>>>>>.

    private void checkForGitConflictMarker() throws IOException {
        skipSpace();
        int markerCount = 0;
        // Looking for the 'ours' marker
        char c;
        while ((c = (char) peek()) == '<' && !isEOFCharacter(c)) {
            read();
            markerCount++;
        }
        if (markerCount == 7) {
            parserResult.addWarning("Found git conflict markers at line %d".formatted(line));
            // Skip 'ours' marker <<<<<<<
            skipLine();
            // Keep skipping lines until we hit the beginning of 'theirs' marker >>>>>>>
            while (peek() != '>' && !isEOFCharacter(peek())) {
                skipLine();
            }
            // Skip 'theirs' marker if we haven't hit EOF already
            if (!isEOFCharacter(peek())) {
                skipLine();
            }
        }
    }

    private void skipLine() throws IOException {
        while (peek() != '\n' && !isEOFCharacter(peek())) {
            read();
        }
        skipOneNewline();
    }

I had to modify the test slightly to pass because the logic I used would continue parsing after the marker, resulting in a parser result with two entries when the expected parser result is zero. I changed it to check whether the warning list contains the git conflict warning.

       assertTrue(parserResult.warnings().contains("Found git conflict markers at line 3"));

[koppor]

Where is the commit? 🎉👀

[HoussemNasri]

I didn't make a commit. I got the inspiration of the solution while working on another PR, so I just made the changes on that PR's branch. However, you can use the code above inside BibTeXParser, and don't forget to set co-authored with me 😁.