# Diagnostic only: find the correct way to expand a .cwl.gz word list.
#
# Two attempts have now produced empty dictionaries:
#   word-list-compress d  -> 0 words   (that is the aspell 0.50 tool)
#   prezip-bin d          -> 4 words   (wrong invocation)
# Guessing a third time is silly. Try every candidate against one known file and
# print the line count for each, so the right one is chosen on evidence.
#
# The reference is en_GB-ise-wo_accents-only.cwl.gz, which should expand to tens
# of thousands of words.
set -x
rm -rf /build/out; mkdir -p /build/out
cd /build/aspell-en-wordlists || { echo NO_WORDLISTS; exit 1; }

F=en_GB-ise-wo_accents-only.cwl.gz
echo "=== the file:"
ls -l "$F"
gunzip -c "$F" | wc -c | sed 's/^/gz_expands_to_bytes=/'
echo "=== first bytes after gunzip:"
gunzip -c "$F" | od -c | head -3

echo "=== what aspell installed that could decompress it:"
ls -l /usr/bin/prezip* /usr/bin/precat /usr/bin/preunzip /usr/bin/word-list-compress 2>/dev/null

echo "=== candidate 1: precat on the .gz directly (what Debian's autobuildhash uses)"
/usr/bin/precat "$F" 2>/dev/null | wc -l | sed 's/^/words=/'

echo "=== candidate 2: preunzip"
gunzip -c "$F" > /tmp/t.cwl
/usr/bin/preunzip < /tmp/t.cwl 2>/dev/null | wc -l | sed 's/^/words=/'

echo "=== candidate 3: prezip-bin d (already tried, for comparison)"
/usr/bin/prezip-bin d < /tmp/t.cwl 2>/dev/null | wc -l | sed 's/^/words=/'

echo "=== candidate 4: prezip d"
/usr/bin/prezip d < /tmp/t.cwl 2>/dev/null | wc -l | sed 's/^/words=/'

echo "=== candidate 5: precat on the uncompressed .cwl"
/usr/bin/precat /tmp/t.cwl 2>/dev/null | wc -l | sed 's/^/words=/'

echo "=== candidate 6: word-list-compress d (the 0.50 tool, for comparison)"
/usr/bin/word-list-compress d < /tmp/t.cwl 2>/dev/null | wc -l | sed 's/^/words=/'

echo "=== whichever gave a large count, show its first lines:"
/usr/bin/precat "$F" 2>/dev/null | head -5
echo "---"
/usr/bin/preunzip < /tmp/t.cwl 2>/dev/null | head -5

echo BUILD_SCRIPT_DONE