Compare commits

...

2 Commits

Author SHA1 Message Date
Xavier Roche
6fb5f71806 entities: harden the generator collision guard and widen test coverage
Review follow-up. The switch keys on the hash alone, so check hash-alone
uniqueness among emitted names (a same-hash/different-len pair would otherwise
slip the old (hash,len) check and surface only as a cryptic duplicate-case
compile error). Also hash the ~93 skipped multi-codepoint names and abort if any
aliases an emitted hash, so "skipped means verbatim" is enforced rather than
assumed on future regens.

Add a runtime sweep of common HTML4 names (copy/reg/trade/mdash/ndash/alpha/beta)
to 01_engine-entities.test: a regression guard against accidental drops and a
generator-vs-consumer hash cross-check on names beyond the handful already
probed.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: Xavier Roche <roche@httrack.com>
2026-06-28 15:23:40 +02:00
Xavier Roche
e65f556a1b Modernize HTML entity decoding to the WHATWG named character references
Regenerate htsentities.h from the WHATWG entities.json (2032 single-codepoint
names) instead of the 1998 HTML 4.0 set (252 names). The dispatch hash moves
from a 32-bit LCG to 64-bit FNV-1a; the generator now aborts on any (hash,len)
collision, so the hash-only switch stays correct without a runtime name compare.
Bump the consumer name-length cap from 10 to 31, the longest name
(CounterClockwiseContourIntegral), or long names would be rejected outright.
Multi-codepoint references (~93 obscure math entities) can't fit the
single-codepoint return and are skipped, left verbatim as before.

Also fix the dead ftp://ftp.unicode.org URLs in htsbasiccharsets.sh.

Closes #443

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: Xavier Roche <roche@httrack.com>
2026-06-28 15:07:04 +02:00
5 changed files with 12251 additions and 1549 deletions

View File

@@ -3,12 +3,12 @@
# Change this to download files
if false; then
echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/ISO8859/8859-*.TXT" | lftp
echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/PC/CP*.TXT" | lftp
echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP*.TXT" | lftp
echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/EBCDIC/CP*.TXT" | lftp
echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MISC/CP*.TXT" | lftp
echo "mget ftp://ftp.unicode.org/Public/MAPPINGS/VENDORS/MISC/KOI8*.TXT" | lftp
echo "mget https://www.unicode.org/Public/MAPPINGS/ISO8859/8859-*.TXT" | lftp
echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/PC/CP*.TXT" | lftp
echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/WINDOWS/CP*.TXT" | lftp
echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MICSFT/EBCDIC/CP*.TXT" | lftp
echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MISC/CP*.TXT" | lftp
echo "mget https://www.unicode.org/Public/MAPPINGS/VENDORS/MISC/KOI8*.TXT" | lftp
rm -f CP932.TXT CP936.TXT CP949.TXT CP950.TXT
fi

View File

@@ -30,12 +30,14 @@ Please visit our Website: http://www.httrack.com
/* Author: Xavier Roche */
/* ------------------------------------------------------------ */
#include <stdint.h>
#include "htscharset.h"
#include "htsencoding.h"
#include "htssafe.h"
/* static int decode_entity(const unsigned int hash, const size_t len);
*/
/* static int decode_entity(const uint64_t hash, const size_t len);
*/
#include "htsentities.h"
/* hexadecimal conversion */
@@ -50,30 +52,31 @@ static int get_hex_value(char c) {
return -1;
}
/* Numerical Recipes,
see <http://en.wikipedia.org/wiki/Linear_congruential_generator> */
#define HASH_PRIME ( 1664525 )
#define HASH_CONST ( 1013904223 )
#define HASH_ADD(HASH, C) do { \
(HASH) *= HASH_PRIME; \
(HASH) += HASH_CONST; \
(HASH) += (C); \
} while(0)
/* 64-bit FNV-1a; must match htsentities.sh, which keys the entity table on it.
*/
#define HASH_INIT 0xcbf29ce484222325ULL
#define HASH_PRIME 0x100000001b3ULL
#define HASH_ADD(HASH, C) \
do { \
(HASH) ^= (unsigned char) (C); \
(HASH) *= HASH_PRIME; \
} while (0)
int hts_unescapeEntitiesWithCharset(const char *src, char *dest, const size_t max, const char *charset) {
size_t i, j, ampStart, ampStartDest;
int uc;
int hex;
unsigned int hash;
uint64_t hash;
assertf(max != 0);
for(i = 0, j = 0, ampStart = (size_t) -1, ampStartDest = 0,
uc = -1, hex = 0, hash = 0 ; src[i] != '\0' ; i++) {
for (i = 0, j = 0, ampStart = (size_t) -1, ampStartDest = 0, uc = -1, hex = 0,
hash = HASH_INIT;
src[i] != '\0'; i++) {
/* start of entity */
if (src[i] == '&') {
ampStart = i;
ampStartDest = j;
hash = 0;
hash = HASH_INIT;
uc = -1;
}
/* inside a potential entity */
@@ -174,14 +177,11 @@ int hts_unescapeEntitiesWithCharset(const char *src, char *dest, const size_t ma
}
/* alphanumerical entity */
else {
/* alphanum and not too far ('&thetasym;' is the longest) */
if (i <= ampStart + 10 &&
(
(src[i] >= '0' && src[i] <= '9')
|| (src[i] >= 'A' && src[i] <= 'Z')
|| (src[i] >= 'a' && src[i] <= 'z')
)
) {
/* alphanum, capped at the longest name
* '&CounterClockwiseContourIntegral;' (31) */
if (i <= ampStart + 31 && ((src[i] >= '0' && src[i] <= '9') ||
(src[i] >= 'A' && src[i] <= 'Z') ||
(src[i] >= 'a' && src[i] <= 'z'))) {
/* compute hash */
HASH_ADD(hash, (unsigned char) src[i]);
} else {

File diff suppressed because it is too large Load Diff

View File

@@ -1,75 +1,92 @@
#!/bin/bash
#
# Regenerate htsentities.h from the WHATWG named character references.
src=html40.txt
url=http://www.w3.org/TR/1998/REC-html40-19980424/html40.txt
set -euo pipefail
src=entities.json
url=https://html.spec.whatwg.org/entities.json
dest=htsentities.h
(
cat <<EOF
/*
-- ${dest} --
FILE GENERATED BY $0, DO NOT MODIFY
# 64-bit FNV-1a of $1, printed as a C constant. Must match the hash in
# htsencoding.c. The offset basis is stored as its wrapped (signed) bit pattern;
# bash arithmetic is 64-bit two's complement, so the result is bit-exact.
fnv1a() {
local s=$1 i c h=$((0xcbf29ce484222325))
for ((i = 0; i < ${#s}; i++)); do
printf -v c '%d' "'${s:i:1}"
h=$(((h ^ (c & 0xff)) * 0x100000001b3))
done
printf '0x%016xULL' "$h"
}
We compute the LCG hash
(see <http://en.wikipedia.org/wiki/Linear_congruential_generator>)
for each entity. We should in theory check using strncmp() that we
actually have the correct entity, but this is actually statistically
not needed.
if [ ! -f "$src" ]; then
curl -fsS "$url" -o "$src"
fi
We may want to do better, but we expect the hash function to be uniform, and
let the compiler be smart enough to optimize the switch (for example by
checking in log2() intervals)
This code has been generated using the evil $0 script.
*/
# Keep ';'-terminated single-codepoint names; the ~93 multi-codepoint refs can't
# fit decode_entity's single-codepoint return and are skipped (left verbatim).
pairs=$(jq -r '
to_entries
| map(select((.key | endswith(";")) and (.value.codepoints | length == 1)))
| sort_by(.key)
| .[] | "\(.key | ltrimstr("&") | rtrimstr(";"))\t\(.value.codepoints[0])"' "$src")
static int decode_entity(const unsigned int hash, const size_t len) {
# Skipped multi-codepoint names, kept to prove none aliases an emitted hash.
skipped=$(jq -r '
to_entries
| map(select((.key | endswith(";")) and (.value.codepoints | length > 1)))
| .[] | .key | ltrimstr("&") | rtrimstr(";")' "$src")
cases=""
emit_hashes=""
while IFS=$'\t' read -r name cp; do
hash=$(fnv1a "$name")
cases+=" /* $name */"$'\n'
cases+=" case $hash:"$'\n'
cases+=" if (len == ${#name}) {"$'\n'
cases+=" return $cp;"$'\n'
cases+=" }"$'\n'
cases+=" break;"$'\n'
emit_hashes+="$hash"$'\n'
done <<<"$pairs"
skip_hashes=""
while IFS= read -r name; do
[ -n "$name" ] && skip_hashes+="$(fnv1a "$name")"$'\n'
done <<<"$skipped"
# The switch keys on the hash alone, so the dispatch is correct only while every
# emitted name hashes uniquely; prove it here, no runtime name compare needed.
dups=$(printf '%s' "$emit_hashes" | sort | uniq -d || true)
if [ -n "$dups" ]; then
echo "FATAL: two entity names share a hash (duplicate switch case); change the hash:" >&2
echo "$dups" >&2
exit 1
fi
# A skipped name colliding with an emitted hash would mis-decode instead of
# staying verbatim; forbid that too.
aliased=$(comm -12 <(printf '%s' "$emit_hashes" | sort -u) <(printf '%s' "$skip_hashes" | sort -u) || true)
if [ -n "$aliased" ]; then
echo "FATAL: a skipped multi-codepoint name aliases an emitted hash:" >&2
echo "$aliased" >&2
exit 1
fi
cat >"$dest" <<EOF
/* GENERATED by $0 from the WHATWG named character references
(${url}). DO NOT EDIT.
Dispatch keys on a 64-bit FNV-1a hash of the entity name; the generator
aborts on any hash collision, so no runtime name compare is needed. */
#include <stdint.h>
static int decode_entity(const uint64_t hash, const size_t len) {
switch(hash) {
EOF
(
if test -f ${src}; then
cat ${src}
else
GET "${url}"
fi
) |
grep -E '^<!ENTITY [a-zA-Z0-9_]' |
sed \
-e 's/<!ENTITY //' -e "s/[[:space:]][[:space:]]*/ /g" \
-e 's/-->$//' \
-e 's/\([^ ]*\) CDATA "&#\([^\"]*\);" -- \(.*\)/\1 \2 \3/' |
(
read -r A
while test -n "$A"; do
ent="${A%% *}"
code=$(echo "$A" | cut -f2 -d' ')
# compute hash
hash=0
i=0
a=1664525
c=1013904223
m="$((1 << 32))"
while test "$i" -lt ${#ent}; do
d="$(echo -n "${ent:${i}:1}" | hexdump -v -e '/1 "%d"')"
hash="$((((hash * a) % (m) + d + c) % (m)))"
i=$((i + 1))
done
echo -e " /* $A */"
echo -e " case ${hash}u:"
echo -e " if (len == ${#ent} /* && strncmp(ent, \"${ent}\") == 0 */) {"
echo -e " return ${code};"
echo -e " }"
echo -e " break;"
# next
read -r A
done
)
cat <<EOF
}
${cases} }
/* unknown */
return -1;
}
EOF
) >${dest}
echo "wrote $dest ($(grep -c '^ case ' "$dest") entities)" >&2

View File

@@ -18,6 +18,21 @@ ent '&amp;' '&'
ent '&lt;&gt;' '<>'
ent '&eacute;' 'é'
# HTML5 names from the WHATWG set
ent '&hellip;' '…'
ent '&bigcup;' ''
# longest name (31 chars) exercises the name-length cap
ent '&CounterClockwiseContourIntegral;' '∳'
# astral codepoint -> 4-byte UTF-8
ent '&Aopf;' '𝔸'
# multi-codepoint refs are skipped at generation, so left verbatim
ent '&fjlig;' '&fjlig;'
# common HTML4 names still decode (regression guard against accidental drops)
ent '&copy;&reg;&trade;' '©®™'
ent '&mdash;&ndash;' '—–'
ent '&alpha;&beta;' 'αβ'
# numeric: decimal and hex
ent '&#65;&#66;' 'AB'
ent '&#x41;' 'A'