Fix the Markdown in the old format v1 of signal-export

fix-signal-export-format1.pl

#!/usr/bin/env perl -w
# 2024-03-12 huy
# Retractively fixes past chats to confirm with current version of signal-export.
#   This is important because signal-export detects duplicates only if the messages match exactly,
#   and the format changes have created duplicates that aren't exactly the same
#   and thus can't be automatically removed.
# Note this script cleans up entries so that the next run of export-signal should detect duplicates but
#   this script doesn't itself remove duplicates.
#
# Details:
# - For transition from format 1 to format 2, adds newlines at end of messages:
#   See https://github.com/carderne/signal-export/commit/4816f1b3480a5785b9fddfa7b44841333ee3f755
# - For transition from format 1 or 2 to format 3, fixes blockquotes:
#   See https://github.com/carderne/signal-export/commit/4816f1b3480a5785b9fddfa7b44841333ee3f755
#   and https://github.com/carderne/signal-export/pull/112/commits/44213b23d4b3b53a9d523e89952ad76f8600de50

# Optionally, specify the timestamp of the first message that was exported using --newlines (v1.8.0+).
# If not defined, then we assume that you never have never run with --newlines.
# WARNING: if there are multiple timestamps that match, you'll have to manually fix the events
#   in the index.md yourself before running this script to disambiguate.
#   But if you do so, be careful not to save the file in an editor that automatically
#     strips trailing whitespace as that will break signal-export's deduplication.
$START_OF_FORMAT2 = '[2024-01-08 10:45]';

# Optionally, specify the timestamp of the first message that was exported using the blockquote
#   fixes from https://github.com/carderne/signal-export/tree/better-quotes
#$START_OF_FORMAT3 = '[2024-03-11 09:00]';


$SRC = './index.md';
$DST = './index.FIXED.md';

open(SRC, $SRC) or die "Can't open $SRC: $!";
open(DST, ">$DST") or die "Can't open $DST: $!";


$format = 1;
$just_saw_date = 0;
$just_saw_blank_line_after_date = 0;
$newline_needed = 0;
$in_blockquote = 0;

while (<SRC>) {
    # Format 1 regime: fix newlines and (best effort) blockquotes
    if ($format == 1) {
        # This is the second line of the blockquote, which is the only line that's properly
        # quoted so we don't need to change anything.
        if ($in_blockquote == 1 && /^> /) {
            $in_blockquote = 2;
            $just_saw_date = 0;
        }
        # If in blockquote, keep quoting until we detect the end of the blockquote
        elsif ($in_blockquote && !/^>$/) {
            # Note that unless this is the second line of the blockquote,
            #   this will match `> blah`, which will be properly converted to `> > blah`
            s/^/> /;

            $just_saw_date = 0;
        } elsif (/^\[(\d{4}-\d{2}-\d{2} \d{2}:\d{2})\] .*/) {
            # Add the newline for the previous message
            print DST "\n" if $newline_needed;
            $newline_needed = 1;

            $just_saw_date = 1;

            # Check for the date marker for the beginning of format 2
            # NOTE: rindex is just efficient way to check for prefix (nothing special)
            if (defined $START_OF_FORMAT2 && rindex($_, $START_OF_FORMAT2, 0) == 0) {
                $format = 2;
                $newline_needed = 0;
            }

            $just_saw_date = 1;
        } else {
            # Check for start of blockquote
            if ($just_saw_date && !$in_blockquote && /^>$/) {
                s/^>//;

                $in_blockquote = 1;
            }
            # If end of blockquote
            # XXX of course, this is ambiguous: what if the quote actually contains a `>`?
            elsif ($in_blockquote && /^>$/) {
                s/^>//;

                $in_blockquote = 0;
            }

            $just_saw_date = 0;
        }

    }
    # Format 2 regime: only fix  blockquotes (best effort)
    elsif ($format = 2) {
        # This is the second line of the blockquote, which is the only line that's properly
        # quoted so we don't need to change anything.
        if ($just_saw_blank_line_after_date && /^> /) {
            $in_blockquote = 2;
            $just_saw_date = 0;
            $just_saw_blank_line_after_date = 0;
        }
        # If in blockquote, keep quoting until the next blank line
        #   XXX This is very ambiguous as the quotation could just have empty line.
        #      but there's nothing we can do.
        #      However, you may be able to detect this discrepancy after the next export
        #      with the proper blockquoting, because the message(s) won't be deduplicated.
        #      and you'll notice messages out of order:
        #        - Find the older message, fix the blockquote manually (or just copy the new version of the message)
        #        - Re-run the export and the dupliates should be removed
        elsif ($in_blockquote && !/^$/) {
            s/^/> /;

            $just_saw_date = 0;
            $just_saw_blank_line_after_date = 0;
        } elsif (/^\[(\d{4}-\d{2}-\d{2} \d{2}:\d{2})\] .*/) {
            $just_saw_date = 1;
            $just_saw_blank_line_after_date = 0;
            $in_blockquote = 0;

            # Check for the date marker for the beginning of format 3
            # (Efficient way to check for prefix)
            if (defined $START_OF_FORMAT3 && rindex($_, $START_OF_FORMAT3, 0) == 0) {
                $format = 3;
            }
        } else {
            # Check for start of blockquote
            if ($just_saw_date && !$in_blockquote && /^$/) {
                $just_saw_blank_line_after_date = 1
            }
            # If end of blockquote
            # XXX of course, this is ambiguous: what if the quote actually contains an empty line
            elsif ($in_blockquote && /^$/) {
                $in_blockquote = 0;
                $just_saw_blank_line_after_date = 0;
            }

            $just_saw_date = 0;
        }
    }

    # Format 3 regime: no change

    print DST;
}