protocol: simpler Curl_getn_scheme runs faster

Iterating unit test 1627 50,000 times show the new version to be 31%
faster on my machine.

- unit1627: add more test strings, In particular three, five and six letter
  non-existing schemes.

- remove scripts/schemetable.c, not used anymore

Closes #22658
This commit is contained in:
Daniel Stenberg 2026-08-24 11:07:17 +02:00
parent c2676bf9e6
commit 2687751471
No known key found for this signature in database
GPG key ID: 5CC908FDB71E12C2
7 changed files with 274 additions and 272 deletions

View file

@ -17,6 +17,7 @@ codespell \
--skip 'RELEASE-NOTES' \
--skip 'scripts/wcurl' \
--skip 'tests/unit/unit1625.c' \
--skip 'tests/unit/unit1627.c' \
--ignore-regex '.*spellchecker:disable-line' \
--ignore-words '.github/scripts/codespell-ignore.words' \
--

View file

@ -35,4 +35,5 @@ extend-exclude = [
"scripts/wcurl",
"tests/data/test*",
"tests/unit/unit1625.c",
"tests/unit/unit1627.c",
]

View file

@ -462,72 +462,199 @@ const struct Curl_scheme Curl_scheme_wss = {
PORT_HTTPS /* defport */
};
static const struct Curl_scheme *two_letter_scheme(const char *scheme)
{
if((Curl_raw_tolower(scheme[0]) == 'w') &&
(Curl_raw_tolower(scheme[1]) == 's'))
return &Curl_scheme_ws;
return NULL;
}
static const struct Curl_scheme *three_letter_scheme(const char *scheme)
{
char s0 = Curl_raw_tolower(scheme[0]);
char s1 = Curl_raw_tolower(scheme[1]);
char s2 = Curl_raw_tolower(scheme[2]);
if(s0 == 'f') {
if(s1 == 't' && s2 == 'p')
return &Curl_scheme_ftp;
}
else if(s0 == 'w') {
if(s1 == 's' && s2 == 's')
return &Curl_scheme_wss;
}
else if(s0 == 's') {
if(s1 == 'c' && s2 == 'p')
return &Curl_scheme_scp;
if(s1 == 'm' && s2 == 'b')
return &Curl_scheme_smb;
}
return NULL;
}
static const struct Curl_scheme *four_letter_scheme(const char *scheme)
{
char s0 = Curl_raw_tolower(scheme[0]);
char s1 = Curl_raw_tolower(scheme[1]);
char s2 = Curl_raw_tolower(scheme[2]);
char s3 = Curl_raw_tolower(scheme[3]);
if(s3 == 'p') {
if(s0 == 'h') {
if(s1 == 't' && s2 == 't')
return &Curl_scheme_http;
}
else if(s0 == 'i') {
if(s1 == 'm' && s2 == 'a')
return &Curl_scheme_imap;
}
else if(s0 == 'l') {
if(s1 == 'd' && s2 == 'a')
return &Curl_scheme_ldap;
}
else if(s0 == 'r') {
if(s1 == 't' && s2 == 's')
return &Curl_scheme_rtsp;
}
else if(s0 == 't') {
if(s1 == 'f' && s2 == 't')
return &Curl_scheme_tftp;
}
else if(s0 == 's') {
if(s1 == 'f' && s2 == 't')
return &Curl_scheme_sftp;
if(s1 == 'm' && s2 == 't')
return &Curl_scheme_smtp;
}
}
else if(s0 == 'f') {
if(s1 == 't' && s2 == 'p' && s3 == 's')
return &Curl_scheme_ftps;
if(s1 == 'i' && s2 == 'l' && s3 == 'e')
return &Curl_scheme_file;
}
else if(s0 == 'm') {
if(s1 == 'q' && s2 == 't' && s3 == 't')
return &Curl_scheme_mqtt;
}
else if(s0 == 'p') {
if(s1 == 'o' && s2 == 'p' && s3 == '3')
return &Curl_scheme_pop3;
}
else if(s0 == 'd') {
if(s1 == 'i' && s2 == 'c' && s3 == 't')
return &Curl_scheme_dict;
}
else if(s0 == 's') {
if(s1 == 'm' && s2 == 'b' && s3 == 's')
return &Curl_scheme_smbs;
}
return NULL;
}
static const struct Curl_scheme *five_letter_scheme(const char *scheme)
{
char s4 = Curl_raw_tolower(scheme[4]);
if(s4 == 's') {
char s0 = Curl_raw_tolower(scheme[0]);
char s1 = Curl_raw_tolower(scheme[1]);
char s2 = Curl_raw_tolower(scheme[2]);
char s3 = Curl_raw_tolower(scheme[3]);
if(s3 == 'p') {
switch(s0) {
case 'h':
if(s1 == 't' && s2 == 't')
return &Curl_scheme_https;
break;
case 'l':
if(s1 == 'd' && s2 == 'a')
return &Curl_scheme_ldaps;
break;
case 'i':
if(s1 == 'm' && s2 == 'a')
return &Curl_scheme_imaps;
break;
case 's':
if(s1 == 'm' && s2 == 't')
return &Curl_scheme_smtps;
break;
default:
break;
}
}
else if(s0 == 'p') {
if(s1 == 'o' && s2 == 'p' && s3 == '3')
return &Curl_scheme_pop3s;
}
else if(s0 == 'm') {
if(s1 == 'q' && s2 == 't' && s3 == 't')
return &Curl_scheme_mqtts;
}
else if(s0 == 's') {
if(s1 == 'o' && s2 == 'c' && s3 == 'k')
return &Curl_scheme_socks;
}
}
return NULL;
}
static const struct Curl_scheme *six_letter_scheme(const char *scheme)
{
char s0 = Curl_raw_tolower(scheme[0]);
switch(s0) {
case 's':
if(curl_strnequal("ocks4", &scheme[1], 5))
return &Curl_scheme_socks4;
if(curl_strnequal("ocks5", &scheme[1], 5))
return &Curl_scheme_socks5;
break;
case 'g':
if(curl_strnequal("opher", &scheme[1], 5))
return &Curl_scheme_gopher;
break;
case 't':
if(curl_strnequal("elnet", &scheme[1], 5))
return &Curl_scheme_telnet;
break;
}
return NULL;
}
static const struct Curl_scheme *seven_letter_scheme(const char *scheme)
{
char s0 = Curl_raw_tolower(scheme[0]);
if(s0 == 's') {
if(curl_strnequal("ocks4a", &scheme[1], 6))
return &Curl_scheme_socks4a;
if(curl_strnequal("ocks5h", &scheme[1], 6))
return &Curl_scheme_socks5h;
}
else if(s0 == 'g') {
if(curl_strnequal("ophers", &scheme[1], 6))
return &Curl_scheme_gophers;
}
return NULL;
}
/* Returns a struct scheme pointer if the name is a known scheme. Check the
->run struct field for non-NULL to figure out if an implementation is
present. */
const struct Curl_scheme *Curl_getn_scheme(const char *scheme, size_t len)
{
/* table generated by schemetable.c:
1. gcc schemetable.c && ./a.out
2. check how small the table gets
3. tweak the hash algorithm, then rerun from 1
4. when the table is good enough
5. copy the table into this source code
6. make sure this function uses the same hash function that worked for
schemetable.c
*/
static const struct Curl_scheme * const all_schemes[59] = { NULL,
&Curl_scheme_pop3, NULL,
&Curl_scheme_smtps,
&Curl_scheme_socks,
&Curl_scheme_socks4,
&Curl_scheme_socks5, NULL, NULL,
&Curl_scheme_gophers,
&Curl_scheme_ws,
&Curl_scheme_sftp,
&Curl_scheme_socks4a,
&Curl_scheme_scp,
&Curl_scheme_rtsp,
&Curl_scheme_dict, NULL, NULL,
&Curl_scheme_gopher, NULL, NULL, NULL,
&Curl_scheme_wss, NULL,
&Curl_scheme_smb, NULL,
&Curl_scheme_ldap,
&Curl_scheme_ldaps,
&Curl_scheme_imap, NULL, NULL, NULL,
&Curl_scheme_imaps,
&Curl_scheme_https,
&Curl_scheme_tftp,
&Curl_scheme_telnet, NULL, NULL, NULL,
&Curl_scheme_file,
&Curl_scheme_smtp, NULL, NULL, NULL, NULL, NULL, NULL, NULL,
&Curl_scheme_ftp,
&Curl_scheme_mqtt, NULL,
&Curl_scheme_socks5h,
&Curl_scheme_http,
&Curl_scheme_pop3s, NULL,
&Curl_scheme_mqtts, NULL,
&Curl_scheme_smbs,
&Curl_scheme_ftps,
typedef const struct Curl_scheme *(*letterfunc)(const char *ptr);
static const letterfunc parse[] = {
two_letter_scheme,
three_letter_scheme,
four_letter_scheme,
five_letter_scheme,
six_letter_scheme,
seven_letter_scheme
};
if(len && (len <= 7)) {
const char *s = scheme;
size_t l = len;
const struct Curl_scheme *h;
unsigned int c = 443;
while(l) {
c <<= 5;
c += (unsigned int)Curl_raw_tolower(*s);
s++;
l--;
}
if(len < 2 || len > 7)
return NULL;
h = all_schemes[c % 59];
if(h && curl_strnequal(scheme, h->name, len) && !h->name[len])
return h;
}
return NULL;
return parse[len - 2](scheme);
}
const struct Curl_scheme *Curl_get_scheme(const char *scheme)

View file

@ -23,7 +23,7 @@
###########################################################################
EXTRA_DIST = coverage.sh completion.pl firefox-db2pem.sh checksrc.pl \
checksrc-all.pl mk-ca-bundle.pl mk-unity.pl schemetable.c cd2nroff nroff2cd \
checksrc-all.pl mk-ca-bundle.pl mk-unity.pl cd2nroff nroff2cd \
cdall cd2cd managen dmaketgz maketgz release-tools.sh verify-release \
cmakelint.sh cmakeopts.sh mdlinkcheck CMakeLists.txt perlcheck.sh \
pythonlint.sh spacecheck.pl randdisable wcurl top-complexity \

View file

@ -1,186 +0,0 @@
/***************************************************************************
* _ _ ____ _
* Project ___| | | | _ \| |
* / __| | | | |_) | |
* | (__| |_| | _ <| |___
* \___|\___/|_| \_\_____|
*
* Copyright (C) Daniel Stenberg, <daniel@haxx.se>, et al.
*
* This software is licensed as described in the file COPYING, which
* you should have received as part of this distribution. The terms
* are also available at https://curl.se/docs/copyright.html.
*
* You may opt to use, copy, modify, merge, publish, distribute and/or sell
* copies of the Software, and permit persons to whom the Software is
* furnished to do so, under the terms of the COPYING file.
*
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
* SPDX-License-Identifier: curl
*
***************************************************************************/
#include <stdio.h>
#include <curl/curl.h>
/*
* Use this tool to generate an updated table for the Curl_getn_scheme_handler
* function in url.c.
*/
static const char * const scheme[] = {
"dict",
"file",
"ftp",
"ftps",
"gopher",
"gophers",
"http",
"https",
"imap",
"imaps",
"ldap",
"ldaps",
"mqtt",
"mqtts",
"pop3",
"pop3s",
"rtsp",
"scp",
"sftp",
"smb",
"smbs",
"smtp",
"smtps",
"socks",
"socks4",
"socks4a",
"socks5",
"socks5h",
"telnet",
"tftp",
"ws",
"wss",
NULL,
};
unsigned int calc(const char *s, int add, int shift)
{
const char *so = s;
unsigned int c = add;
while(*s) {
c <<= shift;
c += *s;
s++;
}
return c;
}
unsigned int num[100];
unsigned int ix[100];
static void showtable(int try, int init, int shift)
{
int nulls = 0;
int i;
for(i = 0; scheme[i]; ++i)
num[i] = calc(scheme[i], init, shift);
for(i = 0; scheme[i]; ++i)
ix[i] = num[i] % try;
printf("/*\n"
" unsigned int c = %d\n"
" while(l) {\n"
" c <<= %d;\n"
" c += Curl_raw_tolower(*s);\n"
" s++;\n"
" l--;\n"
" }\n"
"*/\n",
init, shift);
printf(" static const struct Curl_scheme * const all_schemes[%d] = {", try);
/* generate table */
for(i = 0; i < try; i++) {
int match = 0;
int j;
for(j = 0; scheme[j]; j++) {
if(ix[j] == i) {
printf("\n &Curl_scheme_%s,", scheme[j]);
match = 1;
nulls = 0;
break;
}
}
if(!match)
printf(" NULL,");
}
printf("\n };\n");
}
int main(void)
{
int i;
int try;
int besttry = 9999;
int bestadd = 0;
int bestshift = 0;
int add;
int shift;
for(shift = 0; shift < 8; shift++) {
for(add = 0; add < 999; add++) {
for(i = 0; scheme[i]; ++i) {
unsigned int v = calc(scheme[i], add, shift);
int j;
int badcombo = 0;
for(j = 0; j < i; j++) {
if(num[j] == v) {
#if 0
printf("NOPE: %u is a dupe (%s and %s)\n",
v, scheme[i], scheme[j]);
#endif
badcombo = 1;
break;
}
}
if(badcombo)
break;
num[i] = v;
}
#if 0
for(i = 0; scheme[i].n; ++i) {
printf("%u - %s\n", num[i], scheme[i].n);
}
#endif
/* try different remainders to find smallest possible table */
for(try = 28; try < 199; try++) {
int good = 1;
for(i = 0; scheme[i]; ++i) {
ix[i] = num[i] % try;
}
/* check for dupes */
for(i = 0; scheme[i] && good; ++i) {
int j;
for(j = 0; j < i; j++) {
if(ix[j] == ix[i]) {
good = 0;
break;
}
}
}
if(good) {
if(try < besttry) {
besttry = try;
bestadd = add;
bestshift = shift;
}
break;
}
}
}
}
showtable(besttry, bestadd, bestshift);
}

View file

@ -18,7 +18,7 @@ Curl_get_scheme unit test
<verify>
<stdout mode="text">
199 invokes
547 invokes
</stdout>
</verify>
</testcase>

View file

@ -26,11 +26,13 @@
#include "url.h"
#include "strcase.h"
#define ITERATIONS1627 1 /* edit for performance measurements */
static CURLcode test_unit1627(const char *arg)
{
UNITTEST_BEGIN_SIMPLE
size_t i, j;
size_t i = 0, j = 0, a;
/* existing schemes in different cases */
static const char * const okay[] = {
/* all upper */
@ -48,9 +50,19 @@ static CURLcode test_unit1627(const char *arg)
"imAP", "imaPS", "LDap", "LDAps", "mQTT", "mqtTS", "pOP3", "pOP3s",
"RtsP", "ScP", "SFtP", "Smb", "smBS", "sMTP", "SMTPs",
"TELNEt", "tFTP", "Ws", "wSS",
"SOCKS", "SOCKS4", "SOCKS5", "SOCKS4A", "SOCKS5H",
"socks", "socks4", "socks5", "socks4a", "socks5h",
"Socks", "sOcks4", "soCks5", "socKs4a", "sockS5h",
};
/* non-existing schemes */
static const char * const notokay[] = {
"a", "A", "ht", "tt", "tt", "p+", "TPP", "PPS", "TSP",
"HER", "1CT", "AbG", "LQp", "rtW", "PkY", "xZq", "LmO",
"hyT", "wQA", "dfG", "BvC", "iuY", "ewQ", "dfG", "Jkl",
"NbV", "cXz", "OiU", "SrE", "QaS", "ghJ", "LmN", "VcX", "PoI",
"YtR", "WqA", "DfG", "JkL", "XcV", "NmM", "WeR", "YuI", "PaS",
"a", "A", "htt", "ttp", "httt", "http+", "HTTPP", "HTTPPS", "HTTSP",
"GROPHER", "D1CT", "AbG", "zLQp", "mNrtW", "PkY", "bVcxZq", "LmO",
"iUhyT", "rEwQA", "xSdfG", "nBvC", "pOiuY", "tRewQ", "aSdfG", "hJkl",
@ -64,37 +76,84 @@ static CURLcode test_unit1627(const char *arg)
"AsDf", "GhJk", "LzXc", "VbNm", "qWeR", "tYuI", "oPaS", "dFgH", "jKlZ",
"xCvB", "nMqW", "eRtY", "uIoP", "aSdF", "gHjK", "lZxC", "vBnM", "QwEr",
"TyUi", "OpAs", "DfGh", "JkLz", "XcVb", "NmqW", "ErTy", "UiOp", "AsDf",
"GhJk", "LzXc", "VbNm"
"GhJk", "LzXc", "VbNm",
"aa", "Aa", "htta", "ttpa", "httta", "http+a", "HTTPPa", "HTTPPSa",
"HTTSPa", "GROPHERa", "D1CTa", "AbGa", "zLQpa", "mNrtWa", "PkYa",
"bVcxZqa", "LmOa", "iUhyTa", "rEwQAa", "xSdfGa", "nBvCa", "pOiuYa",
"tRewQa", "aSdfGa", "hJkla", "mNbVa", "cXza", "pOiUa", "yTrEa", "wQaSa",
"dFghJa", "kLmNa", "bVcXa", "zPoIa", "uYtRa", "eWqAa", "sDfGa", "hJkLa",
"zXcVa", "bNmMa", "qWeRa", "tYuIa", "oPaSa", "dFgHa", "jKlZa", "xCvBa",
"nMqWa", "eRtYa", "uIoPa", "aSdFa", "gHjKa", "lZxCa", "vBnMa", "QwEra",
"TyUia", "OpAsa", "DfGha", "JkLza", "XcVba", "NmqWa", "ErTya", "UiOpa",
"AsDfa", "GhJka", "LzXca", "VbNma", "qweRa", "tyuIa", "opaDa", "fghJa",
"klzxa", "cvbna", "mQWa", "ErTya", "UiOpa", "AsDfa", "GhJka", "LzXca",
"VbNma", "QwEra", "TyUia", "OpAsa", "DfGha", "JkLza", "XcVba", "NmqWa",
"ErTya", "UiOpa", "AsDfa", "GhJka", "LzXca", "VbNma", "qWeRa", "tYuIa",
"oPaSa", "dFgHa", "jKlZa", "xCvBa", "nMqWa", "eRtYa", "uIoPa", "aSdFa",
"gHjKa", "lZxCa", "vBnMa", "QwEra", "TyUia", "OpAsa", "DfGha", "JkLza",
"XcVba", "NmqWa", "ErTya", "UiOpa", "AsDfa", "GhJka", "LzXca", "VbNma",
"aab", "Aab", "httab", "ttpab", "htttab", "http+ab", "HTTPPab", "HTTPPSab",
"HTTSPab", "GROPHERab", "D1CTab", "AbGab", "zLQpab", "mNrtWab", "PkYab",
"bVcxZqab", "LmOab", "iUhyTab", "rEwQAab", "xSdfGab", "nBvCab", "pOiuYab",
"tRewQab", "aSdfGab", "hJklab", "mNbVab", "cXzab", "pOiUab", "yTrEab",
"wQaSab", "dFghJab", "kLmNab", "bVcXab", "zPoIab", "uYtRab", "eWqAab",
"sDfGab", "hJkLab", "zXcVab", "bNmMab", "qWeRab", "tYuIab", "oPaSab",
"dFgHab", "jKlZab", "xCvBab", "nMqWab", "eRtYab", "uIoPab", "aSdFab",
"gHjKab", "lZxCab", "vBnMab", "QwErab", "TyUiab", "OpAsab", "DfGhab",
"JkLzab", "XcVbab", "NmqWab", "ErTyab", "UiOpab", "AsDfab", "GhJkab",
"LzXcab", "VbNmab", "qweRab", "tyuIab", "opaDab", "fghJab", "klzxab",
"cvbnab", "mQWab", "ErTyab", "UiOpab", "AsDfab", "GhJkab", "LzXcab",
"VbNmab", "QwErab", "TyUiab", "OpAsab", "DfGhab", "JkLzab", "XcVbab",
"NmqWab", "ErTyab", "UiOpab", "AsDfab", "GhJkab", "LzXcab", "VbNmab",
"qWeRab", "tYuIab", "oPaSab", "dFgHab", "jKlZab", "xCvBab", "nMqWab",
"eRtYab", "uIoPab", "aSdFab", "gHjKab", "lZxCab", "vBnMab", "QwErab",
"TyUiab", "OpAsab", "DfGhab", "JkLzab", "XcVbab", "NmqWab", "ErTyab",
"UiOpab", "AsDfab", "GhJkab", "LzXcab", "VbNmab",
"dictt", "filee", "ftpp", "ftpss", "gopherr", "gopherss", "httpp",
"httpss", "imapp", "imapss", "ldapp", "ldapss", "mqttt", "mqttss",
"pop33", "pop3ss", "rtspp", "scpp", "sftpp", "smbb", "smbss", "smtpp",
"smtpss", "telnett", "tftpp", "wsw", "wsss",
"DIC", "FIL", "FT", "TPS", "GOPHE", "OPHER", "HTT", "TTPS",
"IMA", "APS", "LDA", "APS", "MQT", "TTS", "POP", "P3S",
"RTS", "SC", "SFT", "SM", "BS", "SMT", "TPS",
"TELNE", "TFT", "W", "S",
};
for(i = 0; i < CURL_ARRAYSIZE(okay); i++) {
char buffer[32];
const struct Curl_scheme *get = Curl_get_scheme(okay[i]);
if(get) {
/* verify that we got the correct scheme */
if(!curl_strequal(get->name, okay[i]))
get = NULL;
for(a = 0 ; a < ITERATIONS1627; a++) {
for(i = 0; i < CURL_ARRAYSIZE(okay); i++) {
char buffer[32];
const struct Curl_scheme *get = Curl_get_scheme(okay[i]);
if(get) {
/* verify that we got the correct scheme */
if(!curl_strequal(get->name, okay[i]))
get = NULL;
}
if(!get) {
curl_mprintf("Input: %s, expected okay\n", okay[i]);
break;
}
Curl_strntolower(buffer, okay[i], strlen(okay[i]));
buffer[strlen(okay[i])] = 0;
if(strcmp(buffer, get->name)) {
curl_mprintf("Input: %s is not lowercase: %s\n", buffer, get->name);
break;
}
}
if(!get) {
curl_mprintf("Input: %s, expected okay\n", okay[i]);
break;
}
Curl_strntolower(buffer, okay[i], strlen(okay[i]));
buffer[strlen(okay[i])] = 0;
if(strcmp(buffer, get->name)) {
curl_mprintf("Input: %s is not lowercase: %s\n", buffer, get->name);
break;
}
}
for(j = 0; j < CURL_ARRAYSIZE(notokay); j++) {
const struct Curl_scheme *get = Curl_get_scheme(notokay[j]);
if(get) {
curl_mprintf("Input: %s, expected not okay\n", notokay[j]);
break;
for(j = 0; j < CURL_ARRAYSIZE(notokay); j++) {
const struct Curl_scheme *get = Curl_get_scheme(notokay[j]);
if(get) {
curl_mprintf("Input: %s, expected not okay\n", notokay[j]);
break;
}
}
}
curl_mprintf("%zu invokes\n", i + j);
curl_mprintf("%zu invokes\n", (i + j) * ITERATIONS1627);
if(i != CURL_ARRAYSIZE(okay) ||
j != CURL_ARRAYSIZE(notokay))