X-Git-Url: http://git.ozlabs.org/?p=ccan;a=blobdiff_plain;f=ccan%2Frfc822%2Frfc822.c;h=2d23b9ca79be756ddaef6e6e5df2c94cb3e01255;hp=f5ffd958e767c10ee9602059fac13fa1e8c59bde;hb=a8722345053b7cd860499aa31fd6bb414c120cc8;hpb=f71d6c64922b68fe3a785ec3a9faf093ef9299e1 diff --git a/ccan/rfc822/rfc822.c b/ccan/rfc822/rfc822.c index f5ffd958..2d23b9ca 100644 --- a/ccan/rfc822/rfc822.c +++ b/ccan/rfc822/rfc822.c @@ -5,27 +5,16 @@ #include #include -#include #include +#include +#include #include -#if !HAVE_MEMMEM -void *memmem(const void *haystack, size_t haystacklen, - const void *needle, size_t needlelen) -{ - const char *p, *last; - - p = haystack; - last = p + haystacklen - needlelen; - - do { - if (memcmp(p, needle, needlelen) == 0) - return (void *)p; - } while (p++ <= last); - - return NULL; -} +#ifdef TAL_USE_TALLOC +#include +#else +#include #endif static void (*allocation_failure_hook)(const char *); @@ -57,10 +46,18 @@ void rfc822_set_allocation_failure_handler(void (*h)(const char *)) } \ } while (0) +/* + * No real point doing fancy resizing hashes, when any given mail + * message is unlikely to have more than a fairly small number of + * distinct header types. This should be ample. + */ +#define INDEX_HASH_SIZE 63 + struct rfc822_msg { const char *data, *end; const char *remainder; struct list_head headers; + struct list_head header_index[INDEX_HASH_SIZE]; const char *body; }; @@ -68,6 +65,14 @@ struct rfc822_header { struct bytestring all, rawname, rawvalue; struct bytestring unfolded; struct list_node list; + struct rfc822_header *name_next; +}; + +struct rfc822_headers_of_name { + struct bytestring name; + struct rfc822_header *first; + struct rfc822_header **lastptr; + struct list_node bucket; }; struct rfc822_msg *rfc822_check(const struct rfc822_msg *msg, @@ -76,6 +81,8 @@ struct rfc822_msg *rfc822_check(const struct rfc822_msg *msg, assert(msg); if (!list_check(&msg->headers, abortstr)) return NULL; + if (!tal_check(msg, abortstr)) + return NULL; return (struct rfc822_msg *)msg; } @@ -88,8 +95,9 @@ struct rfc822_msg *rfc822_check(const struct rfc822_msg *msg, struct rfc822_msg *rfc822_start(const void *ctx, const char *p, size_t len) { struct rfc822_msg *msg; + int i; - msg = talloc(ctx, struct rfc822_msg); + msg = tal(ctx, struct rfc822_msg); ALLOC_CHECK(msg, NULL); msg->data = p; @@ -100,6 +108,9 @@ struct rfc822_msg *rfc822_start(const void *ctx, const char *p, size_t len) list_head_init(&msg->headers); + for (i = 0; i < INDEX_HASH_SIZE; i++) + list_head_init(&msg->header_index[i]); + CHECK(msg, "rfc822_free"); - talloc_free(msg); + tal_free(msg); } static struct rfc822_header *next_header_cached(struct rfc822_msg *msg, @@ -137,6 +148,9 @@ static const char *next_line(const char *start, const char *end) return p ? (p + 1) : end; } +static struct rfc822_header *index_header(struct rfc822_msg *msg, + struct rfc822_header *hdr); + static struct rfc822_header *next_header_parse(struct rfc822_msg *msg) { const char *h, *eh, *ev, *colon; @@ -177,7 +191,7 @@ static struct rfc822_header *next_header_parse(struct rfc822_msg *msg) msg->remainder = eh; - hi = talloc_zero(msg, struct rfc822_header); + hi = talz(msg, struct rfc822_header); ALLOC_CHECK(hi, NULL); hi->all = bytestring(h, eh - h); @@ -194,7 +208,7 @@ static struct rfc822_header *next_header_parse(struct rfc822_msg *msg) CHECK(msg, "unfolded = bytestring(raw.ptr, len); } else { - char *unfold = talloc_array(msg, char, len); + char *unfold = tal_arr(msg, char, len); char *p = unfold; ALLOC_CHECK(unfold, bytestring_NULL); @@ -386,13 +400,95 @@ bool rfc822_header_is(struct rfc822_msg *msg, struct rfc822_header *hdr, return hdr_name_eq(hname, bytestring_from_string(name)); } +static unsigned headerhash(struct bytestring name) +{ + /* + * This is stolen from hash_string() in ccan/hash, but adapted + * to add the xtolower() call and use a bytestring + */ + unsigned ret = 0; + size_t i; + + for (i = 0; i < name.len; i++) + ret = (ret << 5) - ret + xtolower(name.ptr[i]); + + return ret % INDEX_HASH_SIZE; +} + +static struct rfc822_headers_of_name *headers_of_name(struct rfc822_msg *msg, + struct bytestring name) +{ + unsigned hash = headerhash(name); + struct rfc822_headers_of_name *hn; + + list_for_each(&msg->header_index[hash], hn, bucket) { + if (hdr_name_eq(hn->name, name)) + return hn; + } + + return NULL; +} + +static struct rfc822_header *index_header(struct rfc822_msg *msg, + struct rfc822_header *hdr) +{ + struct bytestring hname = rfc822_header_raw_name(msg, hdr); + struct rfc822_headers_of_name *hn = headers_of_name(msg, hname); + + if (!hn) { + unsigned hash = headerhash(hname); + + hn = talz(msg, struct rfc822_headers_of_name); + ALLOC_CHECK(hn, NULL); + + hn->name = hname; + hn->first = NULL; + hn->lastptr = &hn->first; + list_add_tail(&msg->header_index[hash], &hn->bucket); + } + + hdr->name_next = NULL; + *(hn->lastptr) = hdr; + hn->lastptr = &hdr->name_next; + return hdr; +} + +struct rfc822_header *rfc822_first_header_of_name(struct rfc822_msg *msg, + const char *name) +{ + struct bytestring namebs = bytestring_from_string(name); + struct rfc822_headers_of_name *hn = headers_of_name(msg, namebs); + struct rfc822_header *hdr; + + if (hn) + return hn->first; + + do { + hdr = next_header_parse(msg); + if (hdr && rfc822_header_is(msg, hdr, name)) + return hdr; + } while (hdr); + + return NULL; +} + struct rfc822_header *rfc822_next_header_of_name(struct rfc822_msg *msg, struct rfc822_header *hdr, const char *name) { + if (!hdr) + return rfc822_first_header_of_name(msg, name); + + if (hdr->name_next) { + assert(rfc822_header_is(msg, hdr->name_next, name)); + return hdr->name_next; + } + do { - hdr = rfc822_next_header(msg, hdr); - } while (hdr && !rfc822_header_is(msg, hdr, name)); + hdr = next_header_parse(msg); + if (hdr && rfc822_header_is(msg, hdr, name)) + return hdr; + } while (hdr); - return hdr; + return NULL; }