--- src/expand.c~ Mon Jul 22 08:59:48 2002 +++ src/expand.c Mon Sep 23 16:15:29 2002 @@ -43,26 +43,34 @@ static uschar *item_table[] = { US"extract", + US"hash", US"if", + US"length", US"lookup", + US"nhash", #ifdef EXIM_PERL US"perl", #endif US"readfile", US"run", US"sg", + US"substr", US"tr" }; enum { EITEM_EXTRACT, + EITEM_HASH, EITEM_IF, + EITEM_LENGTH, EITEM_LOOKUP, + EITEM_NHASH, #ifdef EXIM_PERL EITEM_PERL, #endif EITEM_READFILE, EITEM_RUN, EITEM_SG, + EITEM_SUBSTR, EITEM_TR }; /* Table of operator names, and corresponding switch numbers */ @@ -563,6 +571,190 @@ /************************************************* +* Extract a substring from a string * +*************************************************/ + +/* Perform the ${substr or ${length expansion operations. + +Arguments: + len will contain the length of the returned string + sub the input string + value1 the offset from the start of the input string to the start of + the output string; if negative, count from the right instead. + value2 the length of the output string which is truncated at the right if + value1 is positive or the left if value1 is negative; if value2 + is negative (i.e. unset) then the output string is the right-hand + part of the input if value1 is positive or the left-hand part if + value1 is negative + +Returns: + A pointer to the output string, or NULL if there is an error. + +*/ + +static uschar * +expand_substr (int *len, uschar *sub, int value1, int value2) +{ +int sublen = Ustrlen(sub); + +if (value1 < 0) /* count from right */ + { + value1 += sublen; + + /* If the position is before the start, skip to the start, unless + length is unset, in which case the substring is null. */ + + if (value1 < 0) + { + if (value2 < 0) value2 = 0; else value2 += value1; + value1 = 0; + } + + /* Otherwise an unset length => characters before the value1 */ + + else if (value2 < 0) + { + value2 = value1; + value1 = 0; + } + } + +/* For a non-negative offset, no length means "rest"; just set +it to the maximum. */ + +else if (value2 < 0) value2 = sublen; + +/* Cut the length down to the maximum possible for the offset value, +and get the required characters. */ + +if (sublen < value1 + value2) value2 = sublen - value1; +if (value2 > 0) + { + *len = value2; + return sub + value1; + } +else + { + *len = 0; + return sub; + } +} + + + + +/************************************************* +* Old-style hash of a string * +*************************************************/ + +/* Perform the ${hash expansion operation. + +Arguments: + len will contain the length of the returned string + sub the input string + value1 the length of the output string; if greater or equal to the input + string length then the input string is returned. + value2 the number of hash characters to use, or 26 if negative + +Returns: + A pointer to the output string, or NULL if there is an error. + +*/ + +static uschar * +expand_hash (int *len, uschar *sub, int value1, int value2) +{ +int sublen = Ustrlen(sub); + +if (value2 < 0) value2 = 26; +else if (value2 > Ustrlen(hashcodes)) + { + expand_string_message = + string_sprintf("hash count \"%d\" too big", value2); + return NULL; + } + +/* Calculate the hash text. We know it is shorter than the original +string, so can safely place it in sub[]. */ + +if (value1 < sublen) + { + int c; + int i = 0; + int j = value1; + while ((c = (sub[j])) != 0) + { + int shift = (c + j++) & 7; + sub[i] ^= (c << shift) | (c >> (8-shift)); + if (++i >= value1) i = 0; + } + for (i = 0; i < value1; i++) + sub[i] = hashcodes[(sub[i]) % value2]; + } +else value1 = sublen; + +*len = value1; +return sub; +} + + + + +/************************************************* +* Numeric hash of a string * +*************************************************/ + +/* Perform the ${nhash expansion operation. The first characters of the +string are treated as most important, and get the highest prime numbers. + +Arguments: + len will contain the length of the returned string + sub the input string + value1 the maximum value of the first part of the result + value2 the maximum value of the second part of the result, + or negative to produce only a one-part result + +Returns: + A pointer to the output string, or NULL if there is an error. + +*/ + +static uschar * +expand_nhash (int *len, uschar *sub, int value1, int value2) +{ +uschar *s = sub; +int i = 0; +unsigned long int total = 0; /* no overflow */ + +while (*s != 0) + { + if (i == 0) i = sizeof(prime)/sizeof(int) - 1; + total += prime[i--] * (unsigned int)(*s++); + } + +/* If value2 is unset, just compute one number */ + +if (value2 < 0) + { + s = string_sprintf("%d", total % value1); + } + +/* Otherwise do a div/mod hash */ + +else + { + total = total % (value1 * value2); + s = string_sprintf("%d/%d", total/value2, total % value2); + } + +*len = Ustrlen(s); +return s; +} + + + + +/************************************************* * Find the value of a header or headers * *************************************************/ @@ -2355,6 +2547,73 @@ continue; } + /* Handle "hash", "length", "nhash", "substr" with expanded parameters. */ + + case EITEM_HASH: + case EITEM_LENGTH: + case EITEM_NHASH: + case EITEM_SUBSTR: + { + int i; + int len; + uschar *ret; + int val[2] = { 0, -1 }; + uschar *sub[3]; + + /* "length" takes only 2 arguments whereas the others take 2 or 3. + Ensure that sub[2] is set in the ${length case. */ + + sub[2] = NULL; + switch(read_subs(sub, (name[0] == 'l') ? 2 : 3, 2, &s, skipping)) + { + case 1: /* Treat as 2 */ + case 2: goto EXPAND_FAILED_CURLY; + case 3: goto EXPAND_FAILED; + } + + /* Juggle the arguments if there are only two of them: always move the + string to the last position; and make ${length{n}{str}} equivalent to + ${substr{0}{n}{str}}. See the defaults for val[] above. */ + + if (sub[2] == NULL) + { + sub[2] = sub[1]; + sub[1] = NULL; + if (name[0] == 'l') + { + sub[1] = sub[0]; + sub[0] = NULL; + } + } + + for (i = 0; i < 2; i++) + { + if (sub[i] == NULL) + continue; + val[i] = (int)Ustrtol(sub[i], &ret, 10); + if (*ret != '\0' || (i && val[i] < 0)) + { + expand_string_message = string_sprintf("\"%s\" is not a%s number", + sub[i], i ? " positive" : ""); + goto EXPAND_FAILED; + } + } + + if (name[0] == 'h') + ret = expand_hash(&len, sub[2], val[0], val[1]); + else if (name[0] == 'n') + ret = expand_nhash(&len, sub[2], val[0], val[1]); + else + ret = expand_substr(&len, sub[2], val[0], val[1]); + + if (ret == NULL) + goto EXPAND_FAILED; + else + yield = string_cat(yield, &size, &ptr, ret, len); + + continue; + } + /* Handle global substitution for "sg" - like Perl's s/xxx/yyy/g operator. We have to save the numerical variables and restore them afterwards. */ @@ -2826,7 +3085,8 @@ int value1 = 0; int value2 = -1; int *pn; - int sublen = Ustrlen(sub); + int len; + uschar *ret; if (arg == NULL) { @@ -2873,113 +3133,20 @@ } value1 *= sign; - /* For a text hash, the first argument is the length, and the - second is the number of characters to use, defaulting to 26. */ + /* Perform the actual operation depending on the name. */ if (name[0] == 'h') - { - if (value2 < 0) value2 = 26; - else if (value2 > Ustrlen(hashcodes)) - { - expand_string_message = - string_sprintf("hash uschar count too big in \"%s\"", name); - goto EXPAND_FAILED; - } - - /* Calculate the hash text. We know it is shorter than the original - string, so can safely place it in sub[]. */ - - if (value1 < sublen) - { - int c; - int i = 0; - int j = value1; - while ((c = (sub[j])) != 0) - { - int shift = (c + j++) & 7; - sub[i] ^= (c << shift) | (c >> (8-shift)); - if (++i >= value1) i = 0; - } - for (i = 0; i < value1; i++) - sub[i] = hashcodes[(sub[i]) % value2]; - } - else value1 = sublen; - - yield = string_cat(yield, &size, &ptr, sub, value1); - continue; - } - - /* Numeric hash. The first characters of the string are treated - as most important, and get the highest prime numbers. */ - - if (name[0] == 'n') - { - uschar *s = sub; - int i = 0; - unsigned long int total = 0; /* no overflow */ - - while (*s != 0) - { - if (i == 0) i = sizeof(prime)/sizeof(int) - 1; - total += prime[i--] * (unsigned int)(*s++); - } - - /* If value2 is unset, just compute one number */ - - if (value2 < 0) - { - s = string_sprintf("%d", total % value1); - } - - /* Otherwise do a div/mod hash */ - + ret = expand_hash(&len, sub, value1, value2); + else if (name[0] == 'n') + ret = expand_nhash(&len, sub, value1, value2); else - { - total = total % (value1 * value2); - s = string_sprintf("%d/%d", total/value2, total % value2); - } + ret = expand_substr(&len, sub, value1, value2); - yield = string_cat(yield, &size, &ptr, s, Ustrlen(s)); - continue; - } - - /* Otherwise we are handling a substring or subbit operation, with - value1 the offset and value2 the length. A negative offset positions - from the rhs. (For "length" and "bits" the offset is always 0.) */ - - if (value1 < 0) /* substr only */ - { - value1 += sublen; - - /* If the position is before the start, skip to the start, unless - length is unset, in which case the substring is null. */ - - if (value1 < 0) - { - if (value2 < 0) value2 = 0; else value2 += value1; - value1 = 0; - } - - /* Otherwise an unset length => characters before the value1 */ - - else if (value2 < 0) - { - value2 = value1; - value1 = 0; - } - } - - /* For a non-negative offset, no length means "rest"; just set - it to the maximum. */ - - else if (value2 < 0) value2 = sublen; - - /* For a substring operation, cut length down to maximum possible for - the offset value, and get the required characters. */ + if (ret == NULL) + goto EXPAND_FAILED; + else + yield = string_cat(yield, &size, &ptr, ret, len); - if (sublen < value1 + value2) value2 = sublen - value1; - if (value2 > 0) - yield = string_cat(yield, &size, &ptr, sub + value1, value2); continue; } --- doc/spec.txt~ Mon Jul 22 08:59:53 2002 +++ doc/spec.txt Mon Sep 23 16:22:56 2002 @@ -5686,6 +5686,13 @@ yields '99'. Two successive separators mean that the field between them is empty (for example, the fifth field above). +${hash{}{}} + +${hash{}{}{}} + + These are the same as ${hash_:} and ${hash__:} + below, except that and are expanded before use. + $header_
: or $h_
: Substitute the contents of the named message header line, for example @@ -5734,6 +5741,11 @@ In this case, the expansion is forced to fail if the condition is not true. The available conditions are described in section 11.6 below. +${length{}{}} + + This is the same as ${length_:} below, except that is + expanded before use. + ${lookup{} {} {} {}} ${lookup {} {} {}} @@ -5788,6 +5800,13 @@ ${lookup nisplus {[name=$local_part],passwd.org_dir:gcos} \ {$value}fail} +${nhash{}{}} + +${nhash{}{}{}} + + These are the same as ${nhash_:} and ${nhash__:} + below, except that and are expanded before use. + ${perl{}{}{}...} This item is available only if Exim has been built to include an embedded @@ -5862,6 +5881,13 @@ ${sg{1=A 4=D 3=C}{\N(\d+)=\N}{K\$1=}} yields 'K1=A K4=D K3=C'. + +${substr{}{}} + +${substr{}{}{}} + + These are the same as ${substr_:} and ${substr__:} + below, except that and are expanded before use. ${tr{}{}{}}