@! This file is used to generate pa_hash.c. A perl script merges this
@! file with the output of gperf to produce the hash functions for
@! tag lookup. Lines starting with @! are comments. Lines which do
@! not being with @! are copied straight to the output file. "@begin NAME
@! /REGEX1/ /REGEX2/" means to skip lines in the input until REGEX1 is
@! matched, and then begin saving output under name NAME, and stop when
@! REGEX2 is matched. "@include NAME" inserts the data saved as "NAME".
@! "@SUB NAME SUBREGEX" performs a substitution on the data saved in NAME.
@!
@! Comments and compaints about this poop go to Michael Toy.
@!
@! The following goop extracts the parts we need from the generated output
@! of gperf. We later merge that goop with custom code to generate
@! the tag lookup function.
@!
@begin MACROS /#define/ /^$/
@begin HASH_TABLE /static unsigned [a-z]+ (asso_values|hash_table)/ /};/
@begin HASH_FUNC /register int hval = len;/ /return hval/
@sub HASH_FUNC /return hval \+/hval +=/
@sub HASH_FUNC /str\[/MYLOWER(str[/
@sub HASH_FUNC /]]/])]/
@begin LENGTH_TABLE /static unsigned char lengthtable/ /};/
@begin TAG_TABLE /static struct pa_TagTable *wordlist/ /};/
@!
/*
** This is a generated file, do not edit it. If you need to make changes,
** edit the file pa_hash.template and re-build pa_hash.c on a UNIX machine.
** This whole hacky pile of poop was done by Michael Toy.
*/
#include "pa_parse.h"
@include MACROS
#define MYLOWER(x) TOLOWER(((x) & 0x7f))
/*************************************
* Function: pa_tokenize_tag
*
* Description: This function maps the passed in string
* to one of the valid tag element tokens, or to
* the UNKNOWN token.
*
* Params: Takes a \0 terminated string.
*
* Returns: a 32 bit token to describe this tag element. On error,
* which means it was not passed an unknown tag element string,
* it returns the token P_UNKNOWN.
*
* Performance Notes:
* Profiling on mac revealed this routine as a big (5%) time sink.
* This function was stolen from pa_mdl.c and merged with the perfect
* hashing code and the tag comparison code so it would be flatter (fewer
* function calls) since those are still expensive on 68K and x86 machines.
*************************************/
intn
pa_tokenize_tag(char *str)
{
@include HASH_TABLE
@include LENGTH_TABLE
@include TAG_TABLE
if (str != NULL)
{
int len = strlen(str);
if (len <= MAX_WORD_LENGTH && len >= MIN_WORD_LENGTH)
{
@include HASH_FUNC
@! "hval" now contains hash value
if (hval <= MAX_HASH_VALUE && hval >= MIN_HASH_VALUE)
{
if (len == lengthtable[hval])
{
register char *tag = wordlist[hval].name;
/*
** The following code was stolen from pa_TagEqual,
** again to make this function flatter.
*/
/*
** While not at the end of the string, if they ever differ
** they are not equal. We know "tag" is already lower case.
*/
while ((*tag != '\0')&&(*str != '\0'))
{
if (*tag != (char) TOLOWER(*str))
return(P_UNKNOWN);
tag++;
str++;
}
/*
** One of the strings has ended, if they are both ended, then they
** are equal, otherwise not.
*/
if ((*tag == '\0')&&(*str == '\0'))
return wordlist[hval].id;
}
}
}
}
return(P_UNKNOWN);
}