xref: /netbsd-src/crypto/external/bsd/heimdal/dist/lib/base/bsearch.c (revision afab4e300d3a9fb07dd8c80daf53d0feb3345706)
1 /*	$NetBSD: bsearch.c,v 1.3 2023/06/19 21:41:42 christos Exp $	*/
2 
3 /*
4  * Copyright (c) 2011, Secure Endpoints Inc.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  * - Redistributions of source code must retain the above copyright
12  *   notice, this list of conditions and the following disclaimer.
13  *
14  * - 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 HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
24  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
25  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
26  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
28  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
30  * OF THE POSSIBILITY OF SUCH DAMAGE.
31  *
32  */
33 
34 #include "baselocl.h"
35 
36 #include <sys/types.h>
37 #include <sys/stat.h>
38 #ifdef HAVE_IO_H
39 #include <io.h>
40 #endif
41 #ifdef HAVE_UNISTD_H
42 #include <unistd.h>
43 #endif
44 #include <fcntl.h>
45 #include <ctype.h>
46 #include <stdio.h>
47 #include <stdlib.h>
48 #include <string.h>
49 #ifdef HAVE_STRINGS_H
50 #include <strings.h>
51 #endif
52 #include <errno.h>
53 #include <assert.h>
54 
55 /*
56  * This file contains functions for binary searching flat text in memory
57  * and in text files where each line is a [variable length] record.
58  * Each record has a key and an optional value separated from the key by
59  * unquoted whitespace.  Whitespace in the key, and leading whitespace
60  * for the value, can be quoted with backslashes (but CR and LF must be
61  * quoted in such a way that they don't appear in the quoted result).
62  *
63  * Binary searching a tree are normally a dead simple algorithm.  It
64  * turns out that binary searching flat text with *variable* length
65  * records is... tricky.  There's no indexes to record beginning bytes,
66  * thus any index selected during the search is likely to fall in the
67  * middle of a record.  When deciding to search a left sub-tree one
68  * might fail to find the last record in that sub-tree on account of the
69  * right boundary falling in the middle of it -- the chosen solution to
70  * this makes left sub-tree searches slightly less efficient than right
71  * sub-tree searches.
72  *
73  * If binary searching flat text in memory is tricky, using block-wise
74  * I/O instead is trickier!  But it's necessary in order to support
75  * large files (which we either can't or wouldn't want to read or map
76  * into memory).  Each block we read has to be large enough that the
77  * largest record can fit in it.  And each block might start and/or end
78  * in the middle of a record.  Here it is the right sub-tree searches
79  * that are less efficient than left sub-tree searches.
80  *
81  * bsearch_common() contains the common text block binary search code.
82  *
83  * _bsearch_text() is the interface for searching in-core text.
84  * _bsearch_file() is the interface for block-wise searching files.
85  */
86 
87 struct bsearch_file_handle {
88     int fd;          /* file descriptor */
89     char *cache;     /* cache bytes */
90     char *page;      /* one double-size page worth of bytes */
91     size_t file_sz;  /* file size */
92     size_t cache_sz; /* cache size */
93     size_t page_sz;  /* page size */
94 };
95 
96 /* Find a new-line */
97 static const char *
find_line(const char * buf,size_t i,size_t right)98 find_line(const char *buf, size_t i, size_t right)
99 {
100     if (i == 0)
101 	return &buf[i];
102     for (; i < right; i++) {
103 	if (buf[i] == '\n') {
104 	    if ((i + 1) < right)
105 		return &buf[i + 1];
106 	    return NULL;
107 	}
108     }
109     return NULL;
110 }
111 
112 /*
113  * Common routine for binary searching text in core.
114  *
115  * Perform a binary search of a char array containing a block from a
116  * text file where each line is a record (LF and CRLF supported).  Each
117  * record consists of a key followed by an optional value separated from
118  * the key by whitespace.  Whitespace can be quoted with backslashes.
119  * It's the caller's responsibility to encode/decode keys/values if
120  * quoting is desired; newlines should be encoded such that a newline
121  * does not appear in the result.
122  *
123  * All output arguments are optional.
124  *
125  * Returns 0 if key is found, -1 if not found, or an error code such as
126  * ENOMEM in case of error.
127  *
128  * Inputs:
129  *
130  * @buf          String to search
131  * @sz           Size of string to search
132  * @key          Key string to search for
133  * @buf_is_start True if the buffer starts with a record, false if it
134  *               starts in the middle of a record or if the caller
135  *               doesn't know.
136  *
137  * Outputs:
138  *
139  * @value        Location to store a copy of the value (caller must free)
140  * @location     Record location if found else the location where the
141  *               record should be inserted (index into @buf)
142  * @cmp	         Set to less than or greater than 0 to indicate that a
143  *               key not found would have fit in an earlier or later
144  *               part of a file.  Callers should use this to decide
145  *               whether to read a block to the left or to the right and
146  *               search that.
147  * @loops        Location to store a count of bisections required for
148  *               search (useful for confirming logarithmic performance)
149  */
150 static int
bsearch_common(const char * buf,size_t sz,const char * key,int buf_is_start,char ** value,size_t * location,int * cmp,size_t * loops)151 bsearch_common(const char *buf, size_t sz, const char *key,
152 	       int buf_is_start, char **value, size_t *location,
153 	       int *cmp, size_t *loops)
154 {
155     const char *linep;
156     size_t key_start, key_len; /* key string in buf */
157     size_t val_start, val_len; /* value string in buf */
158     int key_cmp = -1;
159     size_t k;
160     size_t l;    /* left side of buffer for binary search */
161     size_t r;    /* right side of buffer for binary search */
162     size_t rmax; /* right side of buffer for binary search */
163     size_t i;    /* index into buffer, typically in the middle of l and r */
164     size_t loop_count = 0;
165     int ret = -1;
166 
167     if (value)
168 	*value = NULL;
169     if (cmp)
170 	*cmp = 0;
171     if (loops)
172 	*loops = 0;
173 
174     /* Binary search; file should be sorted */
175     for (l = 0, r = rmax = sz, i = sz >> 1; i >= l && i < rmax; loop_count++) {
176 	heim_assert(i < sz, "invalid aname2lname db index");
177 
178 	/* buf[i] is likely in the middle of a line; find the next line */
179 	linep = find_line(buf, i, rmax);
180 	k = linep ? linep - buf : i;
181 	if (linep == NULL || k >= rmax) {
182 	    /*
183 	     * No new line found to the right; search to the left then
184 	     * but don't change rmax (this isn't optimal, but it's
185 	     * simple).
186 	     */
187 	    if (i == l)
188 		break;
189 	    r = i;
190 	    i = l + ((r - l) >> 1);
191 	    continue;
192 	}
193 	i = k;
194 	heim_assert(i >= l && i < rmax, "invalid aname2lname db index");
195 
196 	/* Got a line; check it */
197 
198 	/* Search for and split on unquoted whitespace */
199 	val_start = 0;
200 	for (key_start = i, key_len = 0, val_len = 0, k = i; k < rmax; k++) {
201 	    if (buf[k] == '\\') {
202 		k++;
203 		continue;
204 	    }
205 	    if (buf[k] == '\r' || buf[k] == '\n') {
206 		/* We now know where the key ends, and there's no value */
207 		key_len = k - i;
208 		break;
209 	    }
210 	    if (!isspace((unsigned char)buf[k]))
211 		continue;
212 
213 	    while (k < rmax && isspace((unsigned char)buf[k])) {
214 		key_len = k - i;
215 		k++;
216 	    }
217 	    if (k < rmax)
218 		val_start = k;
219 	    /* Find end of value */
220 	    for (; k < rmax && buf[k] != '\0'; k++) {
221 		if (buf[k] == '\r' || buf[k] == '\n') {
222 		    val_len = k - val_start;
223 		    break;
224 		}
225 	    }
226 	    break;
227 	}
228 
229 	/*
230 	 * The following logic is for dealing with partial buffers,
231 	 * which we use for block-wise binary searches of large files
232 	 */
233 	if (key_start == 0 && !buf_is_start) {
234 	    /*
235 	     * We're at the beginning of a block that might have started
236 	     * in the middle of a record whose "key" might well compare
237 	     * as greater than the key we're looking for, so we don't
238 	     * bother comparing -- we know key_cmp must be -1 here.
239 	     */
240 	    key_cmp = -1;
241 	    break;
242 	}
243 	if ((val_len && buf[val_start + val_len] != '\n') ||
244 	    (!val_len && buf[key_start + key_len] != '\n')) {
245 	    /*
246 	     * We're at the end of a block that ends in the middle of a
247 	     * record whose "key" might well compare as less than the
248 	     * key we're looking for, so we don't bother comparing -- we
249 	     * know key_cmp must be >= 0 but we can't tell.  Our caller
250 	     * will end up reading a double-size block to handle this.
251 	     */
252 	    key_cmp = 1;
253 	    break;
254 	}
255 
256 	key_cmp = strncmp(key, &buf[key_start], key_len);
257 	if (key_cmp == 0 && strlen(key) != key_len)
258 	    key_cmp = 1;
259 	if (key_cmp < 0) {
260 	    /* search left */
261 	    r = rmax = (linep - buf);
262 	    i = l + ((r - l) >> 1);
263 	    if (location)
264 		*location = key_start;
265 	} else if (key_cmp > 0) {
266 	    /* search right */
267 	    if (l == i)
268 		break; /* not found */
269 	    l = i;
270 	    i = l + ((r - l) >> 1);
271 	    if (location)
272 		*location = val_start + val_len;
273 	} else {
274 	    /* match! */
275 	    if (location)
276 		*location = key_start;
277 	    ret = 0;
278 	    if (val_len && value) {
279 		/* Avoid strndup() so we don't need libroken here yet */
280 		if ((*value = malloc(val_len + 1))) {
281                     (void) memcpy(*value, &buf[val_start], val_len);
282                     (*value)[val_len] = '\0';
283                 } else {
284                     ret = errno;
285                 }
286 	    }
287 	    break;
288 	}
289     }
290 
291     if (cmp)
292 	*cmp = key_cmp;
293     if (loops)
294 	*loops = loop_count;
295 
296     return ret;
297 }
298 
299 /*
300  * Binary search a char array containing sorted text records separated
301  * by new-lines (or CRLF).  Each record consists of a key and an
302  * optional value following the key, separated from the key by unquoted
303  * whitespace.
304  *
305  * All output arguments are optional.
306  *
307  * Returns 0 if key is found, -1 if not found, or an error code such as
308  * ENOMEM in case of error.
309  *
310  * Inputs:
311  *
312  * @buf      Char array pointer
313  * @buf_sz   Size of buf
314  * @key      Key to search for
315  *
316  * Outputs:
317  *
318  * @value    Location where to put the value, if any (caller must free)
319  * @location Record location if found else the location where the record
320  *           should be inserted (index into @buf)
321  * @loops    Location where to put a number of loops (or comparisons)
322  *           needed for the search (useful for benchmarking)
323  */
324 int
_bsearch_text(const char * buf,size_t buf_sz,const char * key,char ** value,size_t * location,size_t * loops)325 _bsearch_text(const char *buf, size_t buf_sz, const char *key,
326 	       char **value, size_t *location, size_t *loops)
327 {
328     return bsearch_common(buf, buf_sz, key, 1, value, location, NULL, loops);
329 }
330 
331 #define MAX_BLOCK_SIZE (1024 * 1024)
332 #define DEFAULT_MAX_FILE_SIZE (1024 * 1024)
333 /*
334  * Open a file for binary searching.  The file will be read in entirely
335  * if it is smaller than @max_sz, else a cache of @max_sz bytes will be
336  * allocated.
337  *
338  * Returns 0 on success, else an error number or -1 if the file is empty.
339  *
340  * Inputs:
341  *
342  * @fname   Name of file to open
343  * @max_sz  Maximum size of cache to allocate, in bytes (if zero, default)
344  * @page_sz Page size (must be a power of two, larger than 256, smaller
345  *          than 1MB; if zero use default)
346  *
347  * Outputs:
348  *
349  * @bfh     Handle for use with _bsearch_file() and _bsearch_file_close()
350  * @reads   Number of reads performed
351  */
352 int
_bsearch_file_open(const char * fname,size_t max_sz,size_t page_sz,bsearch_file_handle * bfh,size_t * reads)353 _bsearch_file_open(const char *fname, size_t max_sz, size_t page_sz,
354 		    bsearch_file_handle *bfh, size_t *reads)
355 {
356     bsearch_file_handle new_bfh = NULL;
357     struct stat st;
358     size_t i;
359     int fd;
360     int ret;
361 
362     *bfh = NULL;
363 
364     if (reads)
365 	*reads = 0;
366 
367     fd = open(fname, O_RDONLY);
368     if (fd == -1)
369 	return errno;
370 
371     if (fstat(fd, &st) == -1) {
372 	ret = errno;
373 	goto err;
374     }
375 
376     if (st.st_size == 0) {
377 	ret = -1; /* no data -> no binary search */
378 	goto err;
379     }
380 
381     /* Validate / default arguments */
382     if (max_sz == 0)
383 	max_sz = DEFAULT_MAX_FILE_SIZE;
384     for (i = page_sz; i; i >>= 1) {
385 	/* Make sure page_sz is a power of two */
386 	if ((i % 2) && (i >> 1)) {
387 	    page_sz = 0;
388 	    break;
389 	}
390     }
391     if (page_sz == 0)
392 #ifdef HAVE_STRUCT_STAT_ST_BLKSIZE
393 	page_sz = st.st_blksize;
394 #else
395 	page_sz = 4096;
396 #endif
397     for (i = page_sz; i; i >>= 1) {
398 	/* Make sure page_sz is a power of two */
399 	if ((i % 2) && (i >> 1)) {
400 	    /* Can't happen! Filesystems always use powers of two! */
401 	    page_sz = 4096;
402 	    break;
403 	}
404     }
405     if (page_sz > MAX_BLOCK_SIZE)
406 	page_sz = MAX_BLOCK_SIZE;
407 
408     new_bfh = calloc(1, sizeof (*new_bfh));
409     if (new_bfh == NULL) {
410 	ret = ENOMEM;
411 	goto err;
412     }
413 
414     new_bfh->fd = fd;
415     new_bfh->page_sz = page_sz;
416     new_bfh->file_sz = st.st_size;
417 
418     if (max_sz >= st.st_size) {
419 	/* Whole-file method */
420 	new_bfh->cache = malloc(st.st_size + 1);
421 	if (new_bfh->cache) {
422 	    new_bfh->cache[st.st_size] = '\0';
423 	    new_bfh->cache_sz = st.st_size;
424 	    ret = read(fd, new_bfh->cache, st.st_size);
425 	    if (ret < 0) {
426 		ret = errno;
427 		goto err;
428 	    }
429 	    if (ret != st.st_size) {
430 		ret = EIO; /* XXX ??? */
431 		goto err;
432 	    }
433 	    if (reads)
434 		*reads = 1;
435 	    (void) close(fd);
436 	    new_bfh->fd = -1;
437 	    *bfh = new_bfh;
438 	    return 0;
439 	}
440     }
441 
442     /* Block-size method, or above malloc() failed */
443     new_bfh->page = malloc(new_bfh->page_sz << 1);
444     if (new_bfh->page == NULL) {
445 	/* Can't even allocate a single double-size page! */
446 	ret = ENOMEM;
447 	goto err;
448     }
449 
450     new_bfh->cache_sz = max_sz < st.st_size ? max_sz : st.st_size;
451     new_bfh->cache = malloc(new_bfh->cache_sz);
452     *bfh = new_bfh;
453 
454     /*
455      * malloc() may have failed because we were asking for a lot of
456      * memory, but we may still be able to operate without a cache,
457      * so let's not fail.
458      */
459     if (new_bfh->cache == NULL) {
460 	new_bfh->cache_sz = 0;
461 	return 0;
462     }
463 
464     /* Initialize cache */
465     for (i = 0; i < new_bfh->cache_sz; i += new_bfh->page_sz)
466 	new_bfh->cache[i] = '\0';
467     return 0;
468 
469 err:
470     (void) close(fd);
471     if (new_bfh) {
472 	free(new_bfh->page);
473 	free(new_bfh->cache);
474 	free(new_bfh);
475     }
476     return ret;
477 }
478 
479 /*
480  * Indicate whether the given binary search file handle will be searched
481  * with block-wise method.
482  */
483 void
_bsearch_file_info(bsearch_file_handle bfh,size_t * page_sz,size_t * max_sz,int * blockwise)484 _bsearch_file_info(bsearch_file_handle bfh,
485 		    size_t *page_sz, size_t *max_sz, int *blockwise)
486 {
487     if (page_sz)
488 	*page_sz = bfh->page_sz;
489     if (max_sz)
490 	*max_sz = bfh->cache_sz;
491     if (blockwise)
492 	*blockwise = (bfh->file_sz != bfh->cache_sz);
493 }
494 
495 /*
496  * Close the given binary file search handle.
497  *
498  * Inputs:
499  *
500  * @bfh Pointer to variable containing handle to close.
501  */
502 void
_bsearch_file_close(bsearch_file_handle * bfh)503 _bsearch_file_close(bsearch_file_handle *bfh)
504 {
505     if (!*bfh)
506 	return;
507     if ((*bfh)->fd >= 0)
508 	(void) close((*bfh)->fd);
509     if ((*bfh)->page)
510 	free((*bfh)->page);
511     if ((*bfh)->cache)
512 	free((*bfh)->cache);
513     free(*bfh);
514     *bfh = NULL;
515 }
516 
517 /*
518  * Private function to get a page from a cache.  The cache is a char
519  * array of 2^n - 1 double-size page worth of bytes, where n is the
520  * number of tree levels that the cache stores.  The cache can be
521  * smaller than n implies.
522  *
523  * The page may or may not be valid.  If the first byte of it is NUL
524  * then it's not valid, else it is.
525  *
526  * Returns 1 if page is in cache and valid, 0 if the cache is too small
527  * or the page is invalid.  The page address is output in @buf if the
528  * cache is large enough to contain it regardless of whether the page is
529  * valid.
530  *
531  * Inputs:
532  *
533  * @bfh      Binary search file handle
534  * @level    Level in the tree that we want a page for
535  * @page_idx Page number in the given level (0..2^level - 1)
536  *
537  * Outputs:
538  *
539  * @buf      Set to address of page if the cache is large enough
540  */
541 static int
get_page_from_cache(bsearch_file_handle bfh,size_t level,size_t page_idx,char ** buf)542 get_page_from_cache(bsearch_file_handle bfh, size_t level, size_t page_idx,
543 		    char **buf)
544 {
545     size_t idx = 0;
546     size_t page_sz;
547 
548     page_sz = bfh->page_sz << 1; /* we use double-size pages in the cache */
549 
550     *buf = NULL;
551 
552     /*
553      * Compute index into cache.  The cache is basically an array of
554      * double-size pages.  The first (zeroth) double-size page in the
555      * cache will be the middle page of the file -- the root of the
556      * tree.  The next two double-size pages will be the left and right
557      * pages of the second level in the tree.  The next four double-size
558      * pages will be the four pages at the next level.  And so on for as
559      * many pages as fit in the cache.
560      *
561      * The page index is the number of the page at the given level.  We
562      * then compute (2^level - 1 + page index) * 2page size, check that
563      * we have that in the cache, check that the page has been read (it
564      * doesn't start with NUL).
565      */
566     if (level)
567 	idx = (1 << level) - 1 + page_idx;
568     if (((idx + 1) * page_sz * 2) > bfh->cache_sz)
569 	return 0;
570 
571     *buf = &bfh->cache[idx * page_sz * 2];
572     if (bfh->cache[idx * page_sz * 2] == '\0')
573 	return 0; /* cache[idx] == NUL -> page not loaded in cache */
574     return 1;
575 }
576 
577 /*
578  * Private function to read a page of @page_sz from @fd at offset @off
579  * into @buf, outputing the number of bytes read, which will be the same
580  * as @page_sz unless the page being read is the last page, in which
581  * case the number of remaining bytes in the file will be output.
582  *
583  * Returns 0 on success or an errno value otherwise (EIO if reads are
584  * short).
585  *
586  * Inputs:
587  *
588  * @bfh        Binary search file handle
589  * @level      Level in the binary search tree that we're at
590  * @page_idx   Page "index" at the @level of the tree that we want
591  * @page       Actual page number that we want
592  * want_double Whether we need a page or double page read
593  *
594  * Outputs:
595  *
596  * @buf        Page read or cached
597  * @bytes      Bytes read (may be less than page or double page size in
598  *             the case of the last page, of course)
599  */
600 static int
read_page(bsearch_file_handle bfh,size_t level,size_t page_idx,size_t page,int want_double,const char ** buf,size_t * bytes)601 read_page(bsearch_file_handle bfh, size_t level, size_t page_idx, size_t page,
602 	  int want_double, const char **buf, size_t *bytes)
603 {
604     int ret;
605     off_t off;
606     size_t expected;
607     size_t wanted;
608     char *page_buf;
609 
610     /* Figure out where we're reading and how much */
611     off = page * bfh->page_sz;
612     if (off < 0)
613 	return EOVERFLOW;
614 
615     wanted = bfh->page_sz << want_double;
616     expected = ((bfh->file_sz - off) > wanted) ? wanted : bfh->file_sz - off;
617 
618     if (get_page_from_cache(bfh, level, page_idx, &page_buf)) {
619 	*buf = page_buf;
620 	*bytes = expected;
621 	return 0; /* found in cache */
622     }
623 
624 
625     *bytes = 0;
626     *buf = NULL;
627 
628     /* OK, we have to read a page or double-size page */
629 
630     if (page_buf)
631 	want_double = 1; /* we'll be caching; we cache double-size pages */
632     else
633 	page_buf = bfh->page; /* we won't cache this page */
634 
635     wanted = bfh->page_sz << want_double;
636     expected = ((bfh->file_sz - off) > wanted) ? wanted : bfh->file_sz - off;
637 
638 #ifdef HAVE_PREAD
639     ret = pread(bfh->fd, page_buf, expected, off);
640 #else
641     if (lseek(bfh->fd, off, SEEK_SET) == (off_t)-1)
642 	return errno;
643     ret = read(bfh->fd, page_buf, expected);
644 #endif
645     if (ret < 0)
646 	return errno;
647 
648     if (ret != expected)
649 	return EIO; /* XXX ??? */
650 
651     *buf = page_buf;
652     *bytes = expected;
653     return 0;
654 }
655 
656 /*
657  * Perform a binary search of a file where each line is a record (LF and
658  * CRLF supported).  Each record consists of a key followed by an
659  * optional value separated from the key by whitespace.  Whitespace can
660  * be quoted with backslashes.  It's the caller's responsibility to
661  * encode/decode keys/values if quoting is desired; newlines should be
662  * encoded such that a newline does not appear in the result.
663  *
664  * The search is done with block-wise I/O (i.e., the whole file is not
665  * read into memory).
666  *
667  * All output arguments are optional.
668  *
669  * Returns 0 if key is found, -1 if not found, or an error code such as
670  * ENOMEM in case of error.
671  *
672  * NOTE: We could improve this by not freeing the buffer, instead
673  *       requiring that the caller provide it.  Further, we could cache
674  *       the top N levels of [double-size] pages (2^N - 1 pages), which
675  *       should speed up most searches by reducing the number of reads
676  *       by N.
677  *
678  * Inputs:
679  *
680  * @fd           File descriptor (file to search)
681  * @page_sz      Page size (if zero then the file's st_blksize will be used)
682  * @key          Key string to search for
683  *
684  * Outputs:
685  *
686  * @value        Location to store a copy of the value (caller must free)
687  * @location     Record location if found else the location where the
688  *               record should be inserted (index into @buf)
689  * @loops        Location to store a count of bisections required for
690  *               search (useful for confirming logarithmic performance)
691  * @reads        Location to store a count of pages read during search
692  *               (useful for confirming logarithmic performance)
693  */
694 int
_bsearch_file(bsearch_file_handle bfh,const char * key,char ** value,size_t * location,size_t * loops,size_t * reads)695 _bsearch_file(bsearch_file_handle bfh, const char *key,
696 	       char **value, size_t *location, size_t *loops, size_t *reads)
697 {
698     int ret;
699     const char *buf;
700     size_t buf_sz;
701     size_t page, l, r;
702     size_t my_reads = 0;
703     size_t my_loops_total = 0;
704     size_t my_loops;
705     size_t level;        /* level in the tree */
706     size_t page_idx = 0; /* page number in the tree level */
707     size_t buf_location;
708     int cmp;
709     int buf_ends_in_eol = 0;
710     int buf_is_start = 0;
711 
712     if (reads)
713 	*reads = 0;
714     if (value)
715 	*value = NULL;
716     if (loops)
717 	*loops = 0;
718 
719     /* If whole file is in memory then search that and we're done */
720     if (bfh->file_sz == bfh->cache_sz)
721 	return _bsearch_text(bfh->cache, bfh->cache_sz, key, value, location, loops);
722 
723     /* Else block-wise binary search */
724 
725     l = 0;
726     r = (bfh->file_sz / bfh->page_sz) + 1;
727     for (level = 0, page = r >> 1; page >= l && page < r ; level++) {
728 	ret = read_page(bfh, level, page_idx, page, 0, &buf, &buf_sz);
729 	if (ret != 0)
730 	    return ret;
731 	my_reads++;
732 	if (buf[buf_sz - 1] == '\r' || buf[buf_sz - 1] == '\n')
733 	    buf_ends_in_eol = 1;
734 	else
735 	    buf_ends_in_eol = 0;
736 
737 	buf_is_start = page == 0 ? 1 : 0;
738 	ret = bsearch_common(buf, (size_t)buf_sz, key, buf_is_start,
739 			     value, &buf_location, &cmp, &my_loops);
740 	if (ret > 0)
741 	    return ret;
742 	/* Found or no we update stats */
743 	my_loops_total += my_loops;
744 	if (loops)
745 	    *loops = my_loops_total;
746 	if (reads)
747 	    *reads = my_reads;
748 	if (location)
749 	    *location = page * bfh->page_sz + buf_location;
750 	if (ret == 0)
751 	    return 0; /* found! */
752 	/* Not found */
753 	if (cmp < 0) {
754 	    /* Search left */
755 	    page_idx <<= 1;
756 	    r = page;
757 	    page = l + ((r - l) >> 1);
758 	    continue;
759 	} else {
760 	    /*
761 	     * Search right, but first search the current and next
762 	     * blocks in case that the record we're looking for either
763 	     * straddles the boundary between this and the next record,
764 	     * or in case the record starts exactly at the next page.
765 	     */
766 	    heim_assert(cmp > 0, "cmp > 0");
767 
768 	    if (!buf_ends_in_eol || page == l || page == (r - 1)) {
769 		ret = read_page(bfh, level, page_idx, page, 1, &buf, &buf_sz);
770 		if (ret != 0)
771 		    return ret;
772 		my_reads++;
773 
774 		buf_is_start = page == l ? 1 : 0;
775 
776 		ret = bsearch_common(buf, (size_t)buf_sz, key, buf_is_start,
777 				     value, &buf_location, &cmp, &my_loops);
778 		if (ret > 0)
779 		    return ret;
780 		my_loops_total += my_loops;
781 		if (loops)
782 		    *loops = my_loops_total;
783 		if (reads)
784 		    *reads = my_reads;
785 		if (location)
786 		    *location = page * bfh->page_sz + buf_location;
787 		if (ret == 0)
788 		    return 0;
789 	    }
790 
791 	    /* Oh well, search right */
792 	    if (l == page && r == (l + 1))
793 		break;
794 	    page_idx = (page_idx << 1) + 1;
795 	    l = page;
796 	    page = l + ((r - l) >> 1);
797 	    continue;
798 	}
799     }
800     return -1;
801 }
802 
803 
804 static int
stdb_open(void * plug,const char * dbtype,const char * dbname,heim_dict_t options,void ** db,heim_error_t * error)805 stdb_open(void *plug, const char *dbtype, const char *dbname,
806 	     heim_dict_t options, void **db, heim_error_t *error)
807 {
808     bsearch_file_handle bfh;
809     char *p;
810     int ret;
811 
812     if (error)
813 	*error = NULL;
814     if (dbname == NULL || *dbname == '\0') {
815 	if (error)
816 	    *error = heim_error_create(EINVAL,
817 				       N_("DB name required for sorted-text DB "
818 					  "plugin", ""));
819 	return EINVAL;
820     }
821     p = strrchr(dbname, '.');
822     if (p == NULL || strcmp(p, ".txt") != 0) {
823 	if (error)
824 	    *error = heim_error_create(ENOTSUP,
825 				       N_("Text file (name ending in .txt) "
826 				       "required for sorted-text DB plugin",
827 				       ""));
828 	return ENOTSUP;
829     }
830 
831     ret = _bsearch_file_open(dbname, 0, 0, &bfh, NULL);
832     if (ret)
833 	return ret;
834 
835     *db = bfh;
836     return 0;
837 }
838 
839 static int
stdb_close(void * db,heim_error_t * error)840 stdb_close(void *db, heim_error_t *error)
841 {
842     bsearch_file_handle bfh = db;
843 
844     if (error)
845 	*error = NULL;
846     _bsearch_file_close(&bfh);
847     return 0;
848 }
849 
850 static heim_data_t
stdb_copy_value(void * db,heim_string_t table,heim_data_t key,heim_error_t * error)851 stdb_copy_value(void *db, heim_string_t table, heim_data_t key,
852 	       heim_error_t *error)
853 {
854     bsearch_file_handle bfh = db;
855     const char *k;
856     char *v = NULL;
857     heim_data_t value;
858     int ret;
859 
860     if (error)
861 	*error = NULL;
862 
863     if (table == NULL)
864 	table = HSTR("");
865 
866     if (table != HSTR(""))
867 	return NULL;
868 
869     if (heim_get_tid(key) == HEIM_TID_STRING)
870 	k = heim_string_get_utf8((heim_string_t)key);
871     else
872 	k = (const char *)heim_data_get_ptr(key);
873     ret = _bsearch_file(bfh, k, &v, NULL, NULL, NULL);
874     if (ret == 0 && v == NULL)
875         ret = -1; /* Quiet lint */
876     if (ret != 0) {
877 	if (ret > 0 && error)
878 	    *error = heim_error_create(ret, "%s", strerror(ret));
879 	return NULL;
880     }
881     value = heim_data_create(v, strlen(v));
882     free(v);
883     /* XXX Handle ENOMEM */
884     return value;
885 }
886 
887 struct heim_db_type heim_sorted_text_file_dbtype = {
888     1, stdb_open, NULL, stdb_close, NULL, NULL, NULL, NULL, NULL, NULL,
889     stdb_copy_value, NULL, NULL, NULL
890 };
891