X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=src%2Fcqltransform.c;h=77236b988606e100111a34b261c9ed24ff62b5a6;hb=11dbebdf973d652e486f2b5e457cc46d1478556f;hp=9b2e18d0f460e278e3e0a292997e63f0ec56cf03;hpb=8dde2679aeaae3fbe0cbee6c891ba07249245837;p=yaz-moved-to-github.git diff --git a/src/cqltransform.c b/src/cqltransform.c index 9b2e18d..77236b9 100644 --- a/src/cqltransform.c +++ b/src/cqltransform.c @@ -1,4 +1,4 @@ -/* $Id: cqltransform.c,v 1.20 2006-03-10 17:18:09 mike Exp $ +/* $Id: cqltransform.c,v 1.23 2006-07-05 14:50:16 adam Exp $ Copyright (C) 1995-2005, Index Data ApS Index Data Aps @@ -10,12 +10,24 @@ See the file LICENSE. /** * \file cqltransform.c * \brief Implements CQL transform (CQL to RPN conversion). + * + * Evaluation order of rules: + * + * always + * relation + * structure + * position + * truncation + * index + * relationModifier */ +#include #include #include #include #include +#include struct cql_prop_entry { char *pattern; @@ -209,6 +221,93 @@ int cql_pr_attr(cql_transform_t ct, const char *category, } +static void cql_pr_int (int val, + void (*pr)(const char *buf, void *client_data), + void *client_data) +{ + char buf[21]; /* enough characters to 2^64 */ + sprintf(buf, "%d", val); + (*pr)(buf, client_data); + (*pr)(" ", client_data); +} + + +static int cql_pr_prox(cql_transform_t ct, struct cql_node *mods, + void (*pr)(const char *buf, void *client_data), + void *client_data) +{ + int exclusion = 0; + int distance; /* to be filled in later depending on unit */ + int distance_defined = 0; + int ordered = 0; + int proxrel = 2; /* less than or equal */ + int unit = 2; /* word */ + + while (mods != 0) { + char *name = mods->u.st.index; + char *term = mods->u.st.term; + char *relation = mods->u.st.relation; + + if (!strcmp(name, "distance")) { + distance = strtol(term, (char**) 0, 0); + distance_defined = 1; + if (!strcmp(relation, "=")) { + proxrel = 3; + } else if (!strcmp(relation, ">")) { + proxrel = 5; + } else if (!strcmp(relation, "<")) { + proxrel = 1; + } else if (!strcmp(relation, ">=")) { + proxrel = 4; + } else if (!strcmp(relation, "<=")) { + proxrel = 2; + } else if (!strcmp(relation, "<>")) { + proxrel = 6; + } else { + ct->error = 40; /* Unsupported proximity relation */ + ct->addinfo = xstrdup(relation); + return 0; + } + } else if (!strcmp(name, "ordered")) { + ordered = 1; + } else if (!strcmp(name, "unordered")) { + ordered = 0; + } else if (!strcmp(name, "unit")) { + if (!strcmp(term, "word")) { + unit = 2; + } else if (!strcmp(term, "sentence")) { + unit = 3; + } else if (!strcmp(term, "paragraph")) { + unit = 4; + } else if (!strcmp(term, "element")) { + unit = 8; + } else { + ct->error = 42; /* Unsupported proximity unit */ + ct->addinfo = xstrdup(term); + return 0; + } + } else { + ct->error = 46; /* Unsupported boolean modifier */ + ct->addinfo = xstrdup(name); + return 0; + } + + mods = mods->u.st.modifiers; + } + + if (!distance_defined) + distance = (unit == 2) ? 1 : 0; + + cql_pr_int(exclusion, pr, client_data); + cql_pr_int(distance, pr, client_data); + cql_pr_int(ordered, pr, client_data); + cql_pr_int(proxrel, pr, client_data); + (*pr)("k ", client_data); + cql_pr_int(unit, pr, client_data); + + return 1; +} + /* Returns location of first wildcard character in the `length' * characters starting at `term', or a null pointer of there are * none -- like memchr(). @@ -230,11 +329,16 @@ static const char *wcchar(const char *term, int length) void emit_term(cql_transform_t ct, + struct cql_node *cn, const char *term, int length, void (*pr)(const char *buf, void *client_data), void *client_data) { int i; + const char *ns = cn->u.st.index_uri; + + assert(cn->which == CQL_NODE_ST); + if (length > 0) { if (length > 1 && term[0] == '^' && term[length-1] == '^') @@ -322,6 +426,20 @@ void emit_term(cql_transform_t ct, pr, client_data, 0); } } + if (ns) { + cql_pr_attr_uri(ct, "index", ns, + cn->u.st.index, "serverChoice", + pr, client_data, 16); + } + if (cn->u.st.modifiers) + { + struct cql_node *mod = cn->u.st.modifiers; + for (; mod; mod = mod->u.st.modifiers) + { + cql_pr_attr(ct, "relationModifier", mod->u.st.index, 0, + pr, client_data, 20); + } + } (*pr)("\"", client_data); for (i = 0; iu.st.relation, "eq", pr, client_data, 19); - if (cn->u.st.modifiers) - { - struct cql_node *mod = cn->u.st.modifiers; - for (; mod; mod = mod->u.st.modifiers) - { - cql_pr_attr(ct, "relationModifier", mod->u.st.index, 0, - pr, client_data, 20); - } - } cql_pr_attr(ct, "structure", cn->u.st.relation, 0, pr, client_data, 24); - if (ns) { - cql_pr_attr_uri(ct, "index", ns, - cn->u.st.index, "serverChoice", - pr, client_data, 16); - } if (cn->u.st.relation && !cql_strcmp(cn->u.st.relation, "all")) { emit_wordlist(ct, cn, pr, client_data, "and"); @@ -438,7 +543,7 @@ void cql_transform_r(cql_transform_t ct, } else { - emit_term(ct, cn->u.st.term, strlen(cn->u.st.term), + emit_term(ct, cn, cn->u.st.term, strlen(cn->u.st.term), pr, client_data); } break; @@ -446,6 +551,16 @@ void cql_transform_r(cql_transform_t ct, (*pr)("@", client_data); (*pr)(cn->u.boolean.value, client_data); (*pr)(" ", client_data); + mods = cn->u.boolean.modifiers; + if (!strcmp(cn->u.boolean.value, "prox")) { + if (!cql_pr_prox(ct, mods, pr, client_data)) + return; + } else if (mods) { + /* Boolean modifiers other than on proximity not supported */ + ct->error = 46; /* SRW diag: "Unsupported boolean modifier" */ + ct->addinfo = xstrdup(mods->u.st.index); + return; + } cql_transform_r(ct, cn->u.boolean.left, pr, client_data); cql_transform_r(ct, cn->u.boolean.right, pr, client_data); @@ -498,6 +613,16 @@ int cql_transform_buf(cql_transform_t ct, struct cql_node *cn, info.max = max; info.buf = out; r = cql_transform(ct, cn, cql_buf_write_handler, &info); + if (info.off < 0) { + /* Attempt to write past end of buffer. For some reason, this + SRW diagnostic is deprecated, but it's so perfect for our + purposes that it would be stupid not to use it. */ + char numbuf[30]; + ct->error = YAZ_SRW_TOO_MANY_CHARS_IN_QUERY; + sprintf(numbuf, "%ld", (long) info.max); + ct->addinfo = xstrdup(numbuf); + return -1; + } if (info.off >= 0) info.buf[info.off] = '\0'; return r;