Home | History | Annotate | Line # | Download | only in makemandb
apropos-utils.c revision 1.4.2.4
      1 /*	$NetBSD: apropos-utils.c,v 1.4.2.4 2012/10/30 19:00:37 yamt Exp $	*/
      2 /*-
      3  * Copyright (c) 2011 Abhinav Upadhyay <er.abhinav.upadhyay (at) gmail.com>
      4  * All rights reserved.
      5  *
      6  * This code was developed as part of Google's Summer of Code 2011 program.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  *
     12  * 1. Redistributions of source code must retain the above copyright
     13  *    notice, this list of conditions and the following disclaimer.
     14  * 2. Redistributions in binary form must reproduce the above copyright
     15  *    notice, this list of conditions and the following disclaimer in
     16  *    the documentation and/or other materials provided with the
     17  *    distribution.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
     20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
     21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
     22  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
     23  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
     24  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
     25  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
     26  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
     27  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
     28  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
     29  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     30  * SUCH DAMAGE.
     31  */
     32 
     33 #include <sys/cdefs.h>
     34 __RCSID("$NetBSD: apropos-utils.c,v 1.4.2.4 2012/10/30 19:00:37 yamt Exp $");
     35 
     36 #include <sys/queue.h>
     37 #include <sys/stat.h>
     38 
     39 #include <assert.h>
     40 #include <ctype.h>
     41 #include <err.h>
     42 #include <math.h>
     43 #include <stdio.h>
     44 #include <stdlib.h>
     45 #include <string.h>
     46 #include <util.h>
     47 #include <zlib.h>
     48 
     49 #include "apropos-utils.h"
     50 #include "manconf.h"
     51 #include "mandoc.h"
     52 #include "sqlite3.h"
     53 
     54 typedef struct orig_callback_data {
     55 	void *data;
     56 	int (*callback) (void *, const char *, const char *, const char *,
     57 		const char *, size_t);
     58 } orig_callback_data;
     59 
     60 typedef struct inverse_document_frequency {
     61 	double value;
     62 	int status;
     63 } inverse_document_frequency;
     64 
     65 /* weights for individual columns */
     66 static const double col_weights[] = {
     67 	2.0,	// NAME
     68 	2.00,	// Name-description
     69 	0.55,	// DESCRIPTION
     70 	0.10,	// LIBRARY
     71 	0.001,	//RETURN VALUES
     72 	0.20,	//ENVIRONMENT
     73 	0.01,	//FILES
     74 	0.001,	//EXIT STATUS
     75 	2.00,	//DIAGNOSTICS
     76 	0.05,	//ERRORS
     77 	0.00,	//md5_hash
     78 	1.00	//machine
     79 };
     80 
     81 /*
     82  * lower --
     83  *  Converts the string str to lower case
     84  */
     85 char *
     86 lower(char *str)
     87 {
     88 	assert(str);
     89 	int i = 0;
     90 	char c;
     91 	while (str[i] != '\0') {
     92 		c = tolower((unsigned char) str[i]);
     93 		str[i++] = c;
     94 	}
     95 	return str;
     96 }
     97 
     98 /*
     99 * concat--
    100 *  Utility function. Concatenates together: dst, a space character and src.
    101 * dst + " " + src
    102 */
    103 void
    104 concat(char **dst, const char *src)
    105 {
    106 	concat2(dst, src, strlen(src));
    107 }
    108 
    109 void
    110 concat2(char **dst, const char *src, size_t srclen)
    111 {
    112 	size_t total_len, dst_len;
    113 	assert(src != NULL);
    114 
    115 	/* If destination buffer dst is NULL, then simply strdup the source buffer */
    116 	if (*dst == NULL) {
    117 		*dst = estrdup(src);
    118 		return;
    119 	}
    120 
    121 	dst_len = strlen(*dst);
    122 	/*
    123 	 * NUL Byte and separator space
    124 	 */
    125 	total_len = dst_len + srclen + 2;
    126 
    127 	*dst = erealloc(*dst, total_len);
    128 
    129 	/* Append a space at the end of dst */
    130 	(*dst)[dst_len++] = ' ';
    131 
    132 	/* Now, copy src at the end of dst */
    133 	memcpy(*dst + dst_len, src, srclen + 1);
    134 }
    135 
    136 void
    137 close_db(sqlite3 *db)
    138 {
    139 	sqlite3_close(db);
    140 	sqlite3_shutdown();
    141 }
    142 
    143 /*
    144  * create_db --
    145  *  Creates the database schema.
    146  */
    147 static int
    148 create_db(sqlite3 *db)
    149 {
    150 	const char *sqlstr = NULL;
    151 	char *schemasql;
    152 	char *errmsg = NULL;
    153 
    154 /*------------------------ Create the tables------------------------------*/
    155 
    156 #if NOTYET
    157 	sqlite3_exec(db, "PRAGMA journal_mode = WAL", NULL, NULL, NULL);
    158 #else
    159 	sqlite3_exec(db, "PRAGMA journal_mode = DELETE", NULL, NULL, NULL);
    160 #endif
    161 
    162 	schemasql = sqlite3_mprintf("PRAGMA user_version = %d",
    163 	    APROPOS_SCHEMA_VERSION);
    164 	sqlite3_exec(db, schemasql, NULL, NULL, &errmsg);
    165 	if (errmsg != NULL)
    166 		goto out;
    167 	sqlite3_free(schemasql);
    168 
    169 	sqlstr = "CREATE VIRTUAL TABLE mandb USING fts4(section, name, "
    170 			    "name_desc, desc, lib, return_vals, env, files, "
    171 			    "exit_status, diagnostics, errors, md5_hash UNIQUE, machine, "
    172 			    "compress=zip, uncompress=unzip, tokenize=porter); "	//mandb
    173 			"CREATE TABLE IF NOT EXISTS mandb_meta(device, inode, mtime, "
    174 			    "file UNIQUE, md5_hash UNIQUE, id  INTEGER PRIMARY KEY); "
    175 				//mandb_meta
    176 			"CREATE TABLE IF NOT EXISTS mandb_links(link, target, section, "
    177 			    "machine, md5_hash); ";	//mandb_links
    178 
    179 	sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
    180 	if (errmsg != NULL)
    181 		goto out;
    182 
    183 	sqlstr = "CREATE INDEX IF NOT EXISTS index_mandb_links ON mandb_links "
    184 			"(link); "
    185 			"CREATE INDEX IF NOT EXISTS index_mandb_meta_dev ON mandb_meta "
    186 			"(device, inode); "
    187 			"CREATE INDEX IF NOT EXISTS index_mandb_links_md5 ON mandb_links "
    188 			"(md5_hash);";
    189 	sqlite3_exec(db, sqlstr, NULL, NULL, &errmsg);
    190 	if (errmsg != NULL)
    191 		goto out;
    192 	return 0;
    193 
    194 out:
    195 	warnx("%s", errmsg);
    196 	free(errmsg);
    197 	sqlite3_close(db);
    198 	sqlite3_shutdown();
    199 	return -1;
    200 }
    201 
    202 /*
    203  * zip --
    204  *  User defined Sqlite function to compress the FTS table
    205  */
    206 static void
    207 zip(sqlite3_context *pctx, int nval, sqlite3_value **apval)
    208 {
    209 	int nin;
    210 	long int nout;
    211 	const unsigned char * inbuf;
    212 	unsigned char *outbuf;
    213 
    214 	assert(nval == 1);
    215 	nin = sqlite3_value_bytes(apval[0]);
    216 	inbuf = (const unsigned char *) sqlite3_value_blob(apval[0]);
    217 	nout = nin + 13 + (nin + 999) / 1000;
    218 	outbuf = emalloc(nout);
    219 	compress(outbuf, (unsigned long *) &nout, inbuf, nin);
    220 	sqlite3_result_blob(pctx, outbuf, nout, free);
    221 }
    222 
    223 /*
    224  * unzip --
    225  *  User defined Sqlite function to uncompress the FTS table.
    226  */
    227 static void
    228 unzip(sqlite3_context *pctx, int nval, sqlite3_value **apval)
    229 {
    230 	unsigned int rc;
    231 	unsigned char *outbuf;
    232 	z_stream stream;
    233 
    234 	assert(nval == 1);
    235 	stream.next_in = __UNCONST(sqlite3_value_blob(apval[0]));
    236 	stream.avail_in = sqlite3_value_bytes(apval[0]);
    237 	stream.avail_out = stream.avail_in * 2 + 100;
    238 	stream.next_out = outbuf = emalloc(stream.avail_out);
    239 	stream.zalloc = NULL;
    240 	stream.zfree = NULL;
    241 
    242 	if (inflateInit(&stream) != Z_OK) {
    243 		free(outbuf);
    244 		return;
    245 	}
    246 
    247 	while ((rc = inflate(&stream, Z_SYNC_FLUSH)) != Z_STREAM_END) {
    248 		if (rc != Z_OK ||
    249 		    (stream.avail_out != 0 && stream.avail_in == 0)) {
    250 			free(outbuf);
    251 			return;
    252 		}
    253 		outbuf = erealloc(outbuf, stream.total_out * 2);
    254 		stream.next_out = outbuf + stream.total_out;
    255 		stream.avail_out = stream.total_out;
    256 	}
    257 	if (inflateEnd(&stream) != Z_OK) {
    258 		free(outbuf);
    259 		return;
    260 	}
    261 	outbuf = erealloc(outbuf, stream.total_out);
    262 	sqlite3_result_text(pctx, (const char *) outbuf, stream.total_out, free);
    263 }
    264 
    265 /*
    266  * get_dbpath --
    267  *   Read the path of the database from man.conf and return.
    268  */
    269 char *
    270 get_dbpath(const char *manconf)
    271 {
    272 	TAG *tp;
    273 	char *dbpath;
    274 
    275 	config(manconf);
    276 	tp = gettag("_mandb", 1);
    277 	if (!tp)
    278 		return NULL;
    279 
    280 	if (TAILQ_EMPTY(&tp->entrylist))
    281 		return NULL;
    282 
    283 	dbpath = TAILQ_LAST(&tp->entrylist, tqh)->s;
    284 	return dbpath;
    285 }
    286 
    287 /* init_db --
    288  *   Prepare the database. Register the compress/uncompress functions and the
    289  *   stopword tokenizer.
    290  *	 db_flag specifies the mode in which to open the database. 3 options are
    291  *   available:
    292  *   	1. DB_READONLY: Open in READONLY mode. An error if db does not exist.
    293  *  	2. DB_READWRITE: Open in read-write mode. An error if db does not exist.
    294  *  	3. DB_CREATE: Open in read-write mode. It will try to create the db if
    295  *			it does not exist already.
    296  *  RETURN VALUES:
    297  *		The function will return NULL in case the db does not exist and DB_CREATE
    298  *  	was not specified. And in case DB_CREATE was specified and yet NULL is
    299  *  	returned, then there was some other error.
    300  *  	In normal cases the function should return a handle to the db.
    301  */
    302 sqlite3 *
    303 init_db(int db_flag, const char *manconf)
    304 {
    305 	sqlite3 *db = NULL;
    306 	sqlite3_stmt *stmt;
    307 	struct stat sb;
    308 	int rc;
    309 	int create_db_flag = 0;
    310 
    311 	char *dbpath = get_dbpath(manconf);
    312 	if (dbpath == NULL)
    313 		errx(EXIT_FAILURE, "_mandb entry not found in man.conf");
    314 	/* Check if the database exists or not */
    315 	if (!(stat(dbpath, &sb) == 0 && S_ISREG(sb.st_mode))) {
    316 		/* Database does not exist, check if DB_CREATE was specified, and set
    317 		 * flag to create the database schema
    318 		 */
    319 		if (db_flag != (MANDB_CREATE)) {
    320 			warnx("Missing apropos database. "
    321 			      "Please run makemandb to create it.");
    322 			return NULL;
    323 		}
    324 		create_db_flag = 1;
    325 	}
    326 
    327 	/* Now initialize the database connection */
    328 	sqlite3_initialize();
    329 	rc = sqlite3_open_v2(dbpath, &db, db_flag, NULL);
    330 
    331 	if (rc != SQLITE_OK) {
    332 		warnx("%s", sqlite3_errmsg(db));
    333 		sqlite3_shutdown();
    334 		return NULL;
    335 	}
    336 
    337 	if (create_db_flag && create_db(db) < 0) {
    338 		warnx("%s", "Unable to create database schema");
    339 		goto error;
    340 	}
    341 
    342 	rc = sqlite3_prepare_v2(db, "PRAGMA user_version", -1, &stmt, NULL);
    343 	if (rc != SQLITE_OK) {
    344 		warnx("Unable to query schema version: %s",
    345 		    sqlite3_errmsg(db));
    346 		goto error;
    347 	}
    348 	if (sqlite3_step(stmt) != SQLITE_ROW) {
    349 		sqlite3_finalize(stmt);
    350 		warnx("Unable to query schema version: %s",
    351 		    sqlite3_errmsg(db));
    352 		goto error;
    353 	}
    354 	if (sqlite3_column_int(stmt, 0) != APROPOS_SCHEMA_VERSION) {
    355 		sqlite3_finalize(stmt);
    356 		warnx("Incorrect schema version found. "
    357 		      "Please run makemandb -f.");
    358 		goto error;
    359 	}
    360 	sqlite3_finalize(stmt);
    361 
    362 	sqlite3_extended_result_codes(db, 1);
    363 
    364 	/* Register the zip and unzip functions for FTS compression */
    365 	rc = sqlite3_create_function(db, "zip", 1, SQLITE_ANY, NULL, zip, NULL, NULL);
    366 	if (rc != SQLITE_OK) {
    367 		warnx("Unable to register function: compress: %s",
    368 		    sqlite3_errmsg(db));
    369 		goto error;
    370 	}
    371 
    372 	rc = sqlite3_create_function(db, "unzip", 1, SQLITE_ANY, NULL,
    373                                  unzip, NULL, NULL);
    374 	if (rc != SQLITE_OK) {
    375 		warnx("Unable to register function: uncompress: %s",
    376 		    sqlite3_errmsg(db));
    377 		goto error;
    378 	}
    379 	return db;
    380 
    381 error:
    382 	sqlite3_close(db);
    383 	sqlite3_shutdown();
    384 	return NULL;
    385 }
    386 
    387 /*
    388  * rank_func --
    389  *  Sqlite user defined function for ranking the documents.
    390  *  For each phrase of the query, it computes the tf and idf and adds them over.
    391  *  It computes the final rank, by multiplying tf and idf together.
    392  *  Weight of term t for document d = (term frequency of t in d *
    393  *                                      inverse document frequency of t)
    394  *
    395  *  Term Frequency of term t in document d = Number of times t occurs in d /
    396  *	                                        Number of times t appears in all
    397  *											documents
    398  *
    399  *  Inverse document frequency of t = log(Total number of documents /
    400  *										Number of documents in which t occurs)
    401  */
    402 static void
    403 rank_func(sqlite3_context *pctx, int nval, sqlite3_value **apval)
    404 {
    405 	inverse_document_frequency *idf = sqlite3_user_data(pctx);
    406 	double tf = 0.0;
    407 	const unsigned int *matchinfo;
    408 	int ncol;
    409 	int nphrase;
    410 	int iphrase;
    411 	int ndoc;
    412 	int doclen = 0;
    413 	const double k = 3.75;
    414 	/* Check that the number of arguments passed to this function is correct. */
    415 	assert(nval == 1);
    416 
    417 	matchinfo = (const unsigned int *) sqlite3_value_blob(apval[0]);
    418 	nphrase = matchinfo[0];
    419 	ncol = matchinfo[1];
    420 	ndoc = matchinfo[2 + 3 * ncol * nphrase + ncol];
    421 	for (iphrase = 0; iphrase < nphrase; iphrase++) {
    422 		int icol;
    423 		const unsigned int *phraseinfo = &matchinfo[2 + ncol+ iphrase * ncol * 3];
    424 		for(icol = 1; icol < ncol; icol++) {
    425 
    426 			/* nhitcount: number of times the current phrase occurs in the current
    427 			 *            column in the current document.
    428 			 * nglobalhitcount: number of times current phrase occurs in the current
    429 			 *                  column in all documents.
    430 			 * ndocshitcount:   number of documents in which the current phrase
    431 			 *                  occurs in the current column at least once.
    432 			 */
    433   			int nhitcount = phraseinfo[3 * icol];
    434 			int nglobalhitcount = phraseinfo[3 * icol + 1];
    435 			int ndocshitcount = phraseinfo[3 * icol + 2];
    436 			doclen = matchinfo[2 + icol ];
    437 			double weight = col_weights[icol - 1];
    438 			if (idf->status == 0 && ndocshitcount)
    439 				idf->value += log(((double)ndoc / ndocshitcount))* weight;
    440 
    441 			/* Dividing the tf by document length to normalize the effect of
    442 			 * longer documents.
    443 			 */
    444 			if (nglobalhitcount > 0 && nhitcount)
    445 				tf += (((double)nhitcount  * weight) / (nglobalhitcount * doclen));
    446 		}
    447 	}
    448 	idf->status = 1;
    449 
    450 	/* Final score = (tf * idf)/ ( k + tf)
    451 	 *	Dividing by k+ tf further normalizes the weight leading to better
    452 	 *  results.
    453 	 *  The value of k is experimental
    454 	 */
    455 	double score = (tf * idf->value/ ( k + tf)) ;
    456 	sqlite3_result_double(pctx, score);
    457 	return;
    458 }
    459 
    460 /*
    461  *  run_query --
    462  *  Performs the searches for the keywords entered by the user.
    463  *  The 2nd param: snippet_args is an array of strings providing values for the
    464  *  last three parameters to the snippet function of sqlite. (Look at the docs).
    465  *  The 3rd param: args contains rest of the search parameters. Look at
    466  *  arpopos-utils.h for the description of individual fields.
    467  *
    468  */
    469 int
    470 run_query(sqlite3 *db, const char *snippet_args[3], query_args *args)
    471 {
    472 	const char *default_snippet_args[3];
    473 	char *section_clause = NULL;
    474 	char *limit_clause = NULL;
    475 	char *machine_clause = NULL;
    476 	char *query;
    477 	const char *section;
    478 	char *name;
    479 	const char *name_desc;
    480 	const char *machine;
    481 	const char *snippet;
    482 	const char *name_temp;
    483 	char *slash_ptr;
    484 	char *m = NULL;
    485 	int rc;
    486 	inverse_document_frequency idf = {0, 0};
    487 	sqlite3_stmt *stmt;
    488 
    489 	if (args->machine)
    490 		easprintf(&machine_clause, "AND machine = \'%s\' ", args->machine);
    491 
    492 	/* Register the rank function */
    493 	rc = sqlite3_create_function(db, "rank_func", 1, SQLITE_ANY, (void *)&idf,
    494 	                             rank_func, NULL, NULL);
    495 	if (rc != SQLITE_OK) {
    496 		warnx("Unable to register the ranking function: %s",
    497 		    sqlite3_errmsg(db));
    498 		sqlite3_close(db);
    499 		sqlite3_shutdown();
    500 		exit(EXIT_FAILURE);
    501 	}
    502 
    503 	/* We want to build a query of the form: "select x,y,z from mandb where
    504 	 * mandb match :query [AND (section LIKE '1' OR section LIKE '2' OR...)]
    505 	 * ORDER BY rank DESC..."
    506 	 * NOTES: 1. The portion in square brackets is optional, it will be there
    507 	 * only if the user has specified an option on the command line to search in
    508 	 * one or more specific sections.
    509 	 * 2. I am using LIKE operator because '=' or IN operators do not seem to be
    510 	 * working with the compression option enabled.
    511 	 */
    512 
    513 	if (args->sec_nums) {
    514 		char *temp;
    515 		int i;
    516 
    517 		for (i = 0; i < SECMAX; i++) {
    518 			if (args->sec_nums[i] == 0)
    519 				continue;
    520 			easprintf(&temp, " OR section = \'%d\'", i + 1);
    521 			if (section_clause) {
    522 				concat(&section_clause, temp);
    523 				free(temp);
    524 			} else {
    525 				section_clause = temp;
    526 			}
    527 		}
    528 		if (section_clause) {
    529 			/*
    530 			 * At least one section requested, add glue for query.
    531 			 */
    532 			temp = section_clause;
    533 			/* Skip " OR " before first term. */
    534 			easprintf(&section_clause, " AND (%s)", temp + 4);
    535 			free(temp);
    536 		}
    537 	}
    538 	if (args->nrec >= 0) {
    539 		/* Use the provided number of records and offset */
    540 		easprintf(&limit_clause, " LIMIT %d OFFSET %d",
    541 		    args->nrec, args->offset);
    542 	}
    543 
    544 	if (snippet_args == NULL) {
    545 		default_snippet_args[0] = "";
    546 		default_snippet_args[1] = "";
    547 		default_snippet_args[2] = "...";
    548 		snippet_args = default_snippet_args;
    549 	}
    550 	query = sqlite3_mprintf("SELECT section, name, name_desc, machine,"
    551 	    " snippet(mandb, %Q, %Q, %Q, -1, 40 ),"
    552 	    " rank_func(matchinfo(mandb, \"pclxn\")) AS rank"
    553 	    " FROM mandb"
    554 	    " WHERE mandb MATCH %Q %s "
    555 	    "%s"
    556 	    " ORDER BY rank DESC"
    557 	    "%s",
    558 	    snippet_args[0], snippet_args[1], snippet_args[2], args->search_str,
    559 	    machine_clause ? machine_clause : "",
    560 	    section_clause ? section_clause : "",
    561 	    limit_clause ? limit_clause : "");
    562 
    563 	free(machine_clause);
    564 	free(section_clause);
    565 	free(limit_clause);
    566 
    567 	if (query == NULL) {
    568 		*args->errmsg = estrdup("malloc failed");
    569 		return -1;
    570 	}
    571 	rc = sqlite3_prepare_v2(db, query, -1, &stmt, NULL);
    572 	if (rc == SQLITE_IOERR) {
    573 		warnx("Corrupt database. Please rerun makemandb");
    574 		sqlite3_free(query);
    575 		return -1;
    576 	} else if (rc != SQLITE_OK) {
    577 		warnx("%s", sqlite3_errmsg(db));
    578 		sqlite3_free(query);
    579 		return -1;
    580 	}
    581 
    582 	while (sqlite3_step(stmt) == SQLITE_ROW) {
    583 		section = (const char *) sqlite3_column_text(stmt, 0);
    584 		name_temp = (const char *) sqlite3_column_text(stmt, 1);
    585 		name_desc = (const char *) sqlite3_column_text(stmt, 2);
    586 		machine = (const char *) sqlite3_column_text(stmt, 3);
    587 		snippet = (const char *) sqlite3_column_text(stmt, 4);
    588 		if ((slash_ptr = strrchr(name_temp, '/')) != NULL)
    589 			name_temp = slash_ptr + 1;
    590 		if (machine && machine[0]) {
    591 			m = estrdup(machine);
    592 			easprintf(&name, "%s/%s", lower(m),
    593 				name_temp);
    594 			free(m);
    595 		} else {
    596 			name = estrdup((const char *) sqlite3_column_text(stmt, 1));
    597 		}
    598 
    599 		(args->callback)(args->callback_data, section, name, name_desc, snippet,
    600 			strlen(snippet));
    601 
    602 		free(name);
    603 	}
    604 
    605 	sqlite3_finalize(stmt);
    606 	sqlite3_free(query);
    607 	return *(args->errmsg) == NULL ? 0 : -1;
    608 }
    609 
    610 /*
    611  * callback_html --
    612  *  Callback function for run_query_html. It builds the html output and then
    613  *  calls the actual user supplied callback function.
    614  */
    615 static int
    616 callback_html(void *data, const char *section, const char *name,
    617 	const char *name_desc, const char *snippet, size_t snippet_length)
    618 {
    619 	const char *temp = snippet;
    620 	int i = 0;
    621 	size_t sz = 0;
    622 	int count = 0;
    623 	struct orig_callback_data *orig_data = (struct orig_callback_data *) data;
    624 	int (*callback) (void *, const char *, const char *, const char *,
    625 		const char *, size_t) = orig_data->callback;
    626 
    627 	/* First scan the snippet to find out the number of occurrences of {'>', '<'
    628 	 * '"', '&'}.
    629 	 * Then allocate a new buffer with sufficient space to be able to store the
    630 	 * quoted versions of the special characters {&gt;, &lt;, &quot;, &amp;}.
    631 	 * Copy over the characters from the original snippet to this buffer while
    632 	 * replacing the special characters with their quoted versions.
    633 	 */
    634 
    635 	while (*temp) {
    636 		sz = strcspn(temp, "<>\"&\002\003");
    637 		temp += sz + 1;
    638 		count++;
    639 	}
    640 	size_t qsnippet_length = snippet_length + count * 5;
    641 	char *qsnippet = emalloc(qsnippet_length + 1);
    642 	sz = 0;
    643 	while (*snippet) {
    644 		sz = strcspn(snippet, "<>\"&\002\003");
    645 		if (sz) {
    646 			memcpy(&qsnippet[i], snippet, sz);
    647 			snippet += sz;
    648 			i += sz;
    649 		}
    650 
    651 		switch (*snippet++) {
    652 		case '<':
    653 			memcpy(&qsnippet[i], "&lt;", 4);
    654 			i += 4;
    655 			break;
    656 		case '>':
    657 			memcpy(&qsnippet[i], "&gt;", 4);
    658 			i += 4;
    659 			break;
    660 		case '\"':
    661 			memcpy(&qsnippet[i], "&quot;", 6);
    662 			i += 6;
    663 			break;
    664 		case '&':
    665 			/* Don't perform the quoting if this & is part of an mdoc escape
    666 			 * sequence, e.g. \&
    667 			 */
    668 			if (i && *(snippet - 2) != '\\') {
    669 				memcpy(&qsnippet[i], "&amp;", 5);
    670 				i += 5;
    671 			} else {
    672 				qsnippet[i++] = '&';
    673 			}
    674 			break;
    675 		case '\002':
    676 			memcpy(&qsnippet[i], "<b>", 3);
    677 			i += 3;
    678 			break;
    679 		case '\003':
    680 			memcpy(&qsnippet[i], "</b>", 4);
    681 			i += 4;
    682 			break;
    683 		default:
    684 			break;
    685 		}
    686 	}
    687 	qsnippet[++i] = 0;
    688 	(*callback)(orig_data->data, section, name, name_desc,
    689 		(const char *)qsnippet,	qsnippet_length);
    690 	free(qsnippet);
    691 	return 0;
    692 }
    693 
    694 /*
    695  * run_query_html --
    696  *  Utility function to output query result in HTML format.
    697  *  It internally calls run_query only, but it first passes the output to it's
    698  *  own custom callback function, which preprocess the snippet for quoting
    699  *  inline HTML fragments.
    700  *  After that it delegates the call the actual user supplied callback function.
    701  */
    702 int
    703 run_query_html(sqlite3 *db, query_args *args)
    704 {
    705 	struct orig_callback_data orig_data;
    706 	orig_data.callback = args->callback;
    707 	orig_data.data = args->callback_data;
    708 	const char *snippet_args[] = {"\002", "\003", "..."};
    709 	args->callback = &callback_html;
    710 	args->callback_data = (void *) &orig_data;
    711 	return run_query(db, snippet_args, args);
    712 }
    713 
    714 /*
    715  * callback_pager --
    716  *  A callback similar to callback_html. It overstrikes the matching text in
    717  *  the snippet so that it appears emboldened when viewed using a pager like
    718  *  more or less.
    719  */
    720 static int
    721 callback_pager(void *data, const char *section, const char *name,
    722 	const char *name_desc, const char *snippet, size_t snippet_length)
    723 {
    724 	struct orig_callback_data *orig_data = (struct orig_callback_data *) data;
    725 	char *psnippet;
    726 	const char *temp = snippet;
    727 	int count = 0;
    728 	int i = 0;
    729 	size_t sz = 0;
    730 	size_t psnippet_length;
    731 
    732 	/* Count the number of bytes of matching text. For each of these bytes we
    733 	 * will use 2 extra bytes to overstrike it so that it appears bold when
    734 	 * viewed using a pager.
    735 	 */
    736 	while (*temp) {
    737 		sz = strcspn(temp, "\002\003");
    738 		temp += sz;
    739 		if (*temp == '\003') {
    740 			count += 2 * (sz);
    741 		}
    742 		temp++;
    743 	}
    744 
    745 	psnippet_length = snippet_length + count;
    746 	psnippet = emalloc(psnippet_length + 1);
    747 
    748 	/* Copy the bytes from snippet to psnippet:
    749 	 * 1. Copy the bytes before \002 as it is.
    750 	 * 2. The bytes after \002 need to be overstriked till we encounter \003.
    751 	 * 3. To overstrike a byte 'A' we need to write 'A\bA'
    752 	 */
    753 	while (*snippet) {
    754 		sz = strcspn(snippet, "\002");
    755 		memcpy(&psnippet[i], snippet, sz);
    756 		snippet += sz;
    757 		i += sz;
    758 
    759 		/* Don't change this. Advancing the pointer without reading the byte
    760 		 * is causing strange behavior.
    761 		 */
    762 		if (*snippet == '\002')
    763 			snippet++;
    764 		while (*snippet && *snippet != '\003') {
    765 			psnippet[i++] = *snippet;
    766 			psnippet[i++] = '\b';
    767 			psnippet[i++] = *snippet++;
    768 		}
    769 		if (*snippet)
    770 			snippet++;
    771 	}
    772 
    773 	psnippet[i] = 0;
    774 	(orig_data->callback)(orig_data->data, section, name, name_desc, psnippet,
    775 		psnippet_length);
    776 	free(psnippet);
    777 	return 0;
    778 }
    779 
    780 /*
    781  * run_query_pager --
    782  *  Utility function similar to run_query_html. This function tries to
    783  *  pre-process the result assuming it will be piped to a pager.
    784  *  For this purpose it first calls it's own callback function callback_pager
    785  *  which then delegates the call to the user supplied callback.
    786  */
    787 int
    788 run_query_pager(sqlite3 *db, query_args *args)
    789 {
    790 	struct orig_callback_data orig_data;
    791 	orig_data.callback = args->callback;
    792 	orig_data.data = args->callback_data;
    793 	const char *snippet_args[] = {"\002", "\003", "..."};
    794 	args->callback = &callback_pager;
    795 	args->callback_data = (void *) &orig_data;
    796 	return run_query(db, snippet_args, args);
    797 }
    798