diff -ur bow-20010926/argp/argp-help.c bow-ken/argp/argp-help.c
--- bow-20010926/argp/argp-help.c Tue Mar 25 14:21:16 1997
+++ bow-ken/argp/argp-help.c Fri Dec 14 03:06:12 2001
@@ -27,7 +27,8 @@
#include <string.h>
#include <assert.h>
#include <stdarg.h>
-#include <malloc.h>
+#include <sys/types.h>
+#include <sys/malloc.h>
#include <ctype.h>
#if HAVE_STRINGS_H
diff -ur bow-20010926/argp/strndup.c bow-ken/argp/strndup.c
--- bow-20010926/argp/strndup.c Fri Mar 14 09:36:32 1997
+++ bow-ken/argp/strndup.c Fri Dec 14 03:06:21 2001
@@ -5,7 +5,7 @@
#include <stdio.h>
#include <string.h>
#include <sys/types.h>
-#include <malloc.h>
+#include <sys/malloc.h>
/* Find the length of STRING, but scan at most MAXLEN characters.
If no '\0' terminator is found in that many characters, return MAXLEN. */
diff -ur bow-20010926/barrel.c bow-ken/barrel.c
--- bow-20010926/barrel.c Thu Sep 14 16:47:05 2000
+++ bow-ken/barrel.c Fri Dec 14 02:55:22 2001
@@ -19,7 +19,6 @@
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA */
#include <bow/libbow.h>
-#include <values.h>
static int _bow_barrel_version = -1;
#define BOW_DEFAULT_BARREL_VERSION 3
@@ -73,34 +72,25 @@
return di;
}
-/* Add statistics to the barrel BARREL by indexing all the documents
- found when recursively decending directory DIRNAME. Return the number
- of additional documents indexed. */
-int
-bow_barrel_add_from_text_dir (bow_barrel *barrel,
- const char *dirname,
- const char *except_name,
- const char *classname)
-{
- int text_file_count, binary_file_count;
+struct index_file_context {
+ const char *except_name;
+ bow_barrel *barrel;
int class;
+ int text_file_count;
+ int binary_file_count;
+};
+#define CONTEXT ((struct index_file_context*)context)
-#ifdef VPC_ONLY
- /* Used when we are building a class barrel */
- bow_cdoc cdoc;
- bow_cdoc *cdocp;
- int word_count = 0;
- int di;
-
- /* Function used to build a multinomial vpc barrel without building a
- * document barrel */
- int class_barrel_index_file (const char *filename, void *context)
+ int barrel_index_file (const char *filename, void *context)
{
FILE *fp;
+ bow_cdoc cdoc;
+ bow_cdoc *cdocp;
+ int di; /* a document index */
int num_words;
/* If the filename matches the exception name, return immediately. */
- if (except_name && !strcmp (filename, except_name))
+ if (CONTEXT->except_name && !strcmp (filename, CONTEXT->except_name))
return 0;
if (!(fp = fopen (filename, "r")))
@@ -111,34 +101,68 @@
}
if (bow_fp_is_text (fp))
{
+ /* The file contains text; snarf the words and put them in
+ the WI2DVF map. */
+ cdoc.type = bow_doc_train;
+ cdoc.class = CONTEXT->class;
+ /* Set to one so bow_infogain_per_wi_new() works correctly
+ by default. */
+ cdoc.prior = 1.0f;
+ assert (cdoc.class >= 0);
+ cdoc.filename = strdup (filename);
+ assert (cdoc.filename);
+ cdoc.class_probs = NULL;
+ /* Add the CDOC to CDOCS, and determine the "index" of this
+ document. */
+ di = bow_array_append (CONTEXT->barrel->cdocs, &cdoc);
/* Add all the words in this document. */
- num_words = bow_wi2dvf_add_di_text_fp (&(barrel->wi2dvf), di, fp,
+ num_words = bow_wi2dvf_add_di_text_fp (&(CONTEXT->barrel->wi2dvf), di, fp,
filename);
- word_count += num_words;
- text_file_count++;
+ /* Fill in the new CDOC's idea of WORD_COUNT */
+ cdocp = bow_array_entry_at_index (CONTEXT->barrel->cdocs, di);
+ cdocp->word_count = num_words;
+ CONTEXT->text_file_count++;
}
else
{
bow_verbosify (bow_progress,
"\nFile `%s' skipped because not text\n",
filename);
- binary_file_count++;
+ CONTEXT->binary_file_count++;
}
fclose (fp);
bow_verbosify (bow_progress,
"\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b\b"
"%6d : %8d",
- text_file_count, bow_num_words ());
+ CONTEXT->text_file_count, bow_num_words ());
return 1;
}
-#endif
- int barrel_index_file (const char *filename, void *context)
+/* Add statistics to the barrel BARREL by indexing all the documents
+ found when recursively decending directory DIRNAME. Return the number
+ of additional documents indexed. */
+int
+bow_barrel_add_from_text_dir (bow_barrel *barrel,
+ const char *dirname,
+ const char *except_name,
+ const char *classname)
+{
+ struct index_file_context context;
+
+#ifdef VPC_ONLY
+ /* Used when we are building a class barrel */
+ bow_cdoc cdoc;
+ bow_cdoc *cdocp;
+ int word_count = 0;
+ int di;
+
+
+
+ /* Function used to build a multinomial vpc barrel without building a
+ * document barrel */
+ int class_barrel_index_file (const char *filename, void *context)
{
FILE *fp;
- bow_cdoc cdoc;
- bow_cdoc *cdocp;
- int di; /* a document index */
int num_words;
/* If the filename matches the exception name, return immediately. */
@@ -153,26 +177,10 @@
}
if (bow_fp_is_text (fp))
{
- /* The file contains text; snarf the words and put them in
- the WI2DVF map. */
- cdoc.type = bow_doc_train;
- cdoc.class = class;
- /* Set to one so bow_infogain_per_wi_new() works correctly
- by default. */
- cdoc.prior = 1.0f;
- assert (cdoc.class >= 0);
- cdoc.filename = strdup (filename);
- assert (cdoc.filename);
- cdoc.class_probs = NULL;
- /* Add the CDOC to CDOCS, and determine the "index" of this
- document. */
- di = bow_array_append (barrel->cdocs, &cdoc);
/* Add all the words in this document. */
num_words = bow_wi2dvf_add_di_text_fp (&(barrel->wi2dvf), di, fp,
filename);
- /* Fill in the new CDOC's idea of WORD_COUNT */
- cdocp = bow_array_entry_at_index (barrel->cdocs, di);
- cdocp->word_count = num_words;
+ word_count += num_words;
text_file_count++;
}
else
@@ -189,17 +197,19 @@
text_file_count, bow_num_words ());
return 1;
}
+#endif
if (!(barrel->classnames))
barrel->classnames = bow_int4str_new (0);
- class = bow_str2int (barrel->classnames, classname);
+ context.class = bow_str2int (barrel->classnames, classname);
+ context.barrel = barrel;
#ifdef VPC_ONLY
/* If we are building a class barrel, make one cdoc per class */
if (barrel->is_vpc)
{
cdoc.type = bow_doc_train;
- cdoc.class = class;
+ cdoc.class = context.class;
cdoc.filename = strdup (classname);
cdoc.word_count = 0; /* Number of documents in this class */
cdoc.normalizer = -1.0f;
@@ -215,7 +225,7 @@
bow_verbosify (bow_progress,
"Gathering stats... files : unique-words :: "
" ");
- text_file_count = binary_file_count = 0;
+ context.text_file_count = context.binary_file_count = 0;
#ifdef VPC_ONLY
/* Call our special function for building a class barrel rather than
* a document barrel. When finished, set the priors and word counts */
@@ -228,12 +238,12 @@
}
else
#endif
- bow_map_filenames_from_dir (barrel_index_file, 0, dirname, "");
+ bow_map_filenames_from_dir (barrel_index_file, &context, dirname, "");
bow_verbosify (bow_progress, "\n");
- if (binary_file_count > text_file_count)
+ if (context.binary_file_count > context.text_file_count)
bow_verbosify (bow_quiet,
"Found mostly binary files, which were ignored.\n");
- return text_file_count;
+ return context.text_file_count;
}
/* Call this on a vector-per-document barrel to set the CDOC->PRIOR's
@@ -356,21 +366,7 @@
float *wi2ig;
int wi2ig_size;
int wi, i;
- struct wiig_list_entry {
- float ig;
- int wi;
- } *wiig_list;
- /* For sorting the above entries. */
- int compare_wiig_list_entry (const void *e1, const void *e2)
- {
- if (((struct wiig_list_entry*)e1)->ig >
- ((struct wiig_list_entry*)e2)->ig)
- return -1;
- else if (((struct wiig_list_entry*)e1)->ig ==
- ((struct wiig_list_entry*)e2)->ig)
- return 0;
- else return 1;
- }
+ wiig *wiig_list;
if (num_words_to_keep == 0)
return;
@@ -383,15 +379,15 @@
/* Make a list of the info gain numbers paired with their WI's,
in prepartion for sorting. */
- wiig_list = alloca (sizeof (struct wiig_list_entry) * wi2ig_size);
+ wiig_list = alloca (sizeof (wiig) * wi2ig_size);
for (wi = 0; wi < wi2ig_size; wi++)
{
wiig_list[wi].wi = wi;
wiig_list[wi].ig = wi2ig[wi];
}
/* Sort the list */
- qsort (wiig_list, wi2ig_size, sizeof (struct wiig_list_entry),
- compare_wiig_list_entry);
+ qsort (wiig_list, wi2ig_size, sizeof (wiig),
+ wiig_compare);
num_words_to_keep = MIN (num_words_to_keep, wi2ig_size);
@@ -1128,6 +1124,7 @@
bow_dv *dv;
int dvi;
};
+#undef CONTEXT
#define CONTEXT ((struct bow_barrel_iterator_context*)context)
static void
diff -ur bow-20010926/bow/libbow.h bow-ken/bow/libbow.h
--- bow-20010926/bow/libbow.h Mon Feb 19 14:50:43 2001
+++ bow-ken/bow/libbow.h Fri Dec 14 03:06:29 2001
@@ -37,7 +37,7 @@
#include <sys/types.h> /* for netinet/in.h on SunOS */
#include <sys/stat.h>
#include <netinet/in.h> /* for machine-independent byte-order */
-#include <malloc.h> /* for malloc() and friends. */
+#include <sys/malloc.h> /* for malloc() and friends. */
#include <stdlib.h> /* For malloc() etc. on DEC Alpha */
#include <string.h> /* for strlen() on DEC Alpha */
#include <limits.h> /* for PATH_MAX and SHRT_MAX and friends */
@@ -2226,5 +2226,24 @@
} bow_flex_type;
extern bow_flex_type bow_flex_option;
+
+typedef struct _wiig {
+ int wi;
+ float ig;
+} wiig;
+
+struct _doc_hit {
+ int di;
+ float score;
+ const char **terms;
+ int terms_count;
+};
+
+/* For sorting the above entries. */
+int wiig_compare (const void *wiig1, const void *wiig2);
+int compare_doc_hits (struct _doc_hit *hit1, struct _doc_hit *hit2);
+
+#define sqrtf sqrt
+#define MAXDOUBLE DBL_MAX
#endif /* __libbow_h_INCLUDE */
diff -ur bow-20010926/bow/svm.h bow-ken/bow/svm.h
--- bow-20010926/bow/svm.h Fri May 19 13:48:31 2000
+++ bow-ken/bow/svm.h Fri Dec 14 02:23:04 2001
@@ -22,7 +22,6 @@
#include <math.h>
#include <string.h>
-#include <values.h>
#include <time.h>
#include <signal.h>
#include <sys/times.h>
diff -ur bow-20010926/crossbow.c bow-ken/crossbow.c
--- bow-20010926/crossbow.c Tue May 23 17:08:26 2000
+++ bow-ken/crossbow.c Fri Dec 14 02:34:57 2001
@@ -971,9 +971,11 @@
}
}
-void
-crossbow_classify_docs_in_dirname (const char *dirname, int verbose)
-{
+struct classify_filename_context {
+ int verbose;
+};
+#define CONTEXT ((struct classify_filename_context *)context)
+
int classify_filename (const char *filename, void *context)
{
crossbow_doc doc;
@@ -997,11 +999,16 @@
doc.cis_size = -1;
doc.cis = NULL;
((((crossbow_method*)bow_argp_method)->classify_doc)
- (&doc, verbose, stdout));
+ (&doc, CONTEXT->verbose, stdout));
return 0;
}
- bow_map_filenames_from_dir (classify_filename, NULL, dirname, "");
+void
+crossbow_classify_docs_in_dirname (const char *dirname, int verbose)
+{
+ struct classify_filename_context context;
+ context.verbose = verbose;
+ bow_map_filenames_from_dir (classify_filename, &context, dirname, "");
}
diff -ur bow-20010926/dirichlet.c bow-ken/dirichlet.c
--- bow-20010926/dirichlet.c Mon Dec 18 09:38:53 2000
+++ bow-ken/dirichlet.c Fri Dec 14 03:05:59 2001
@@ -1,6 +1,6 @@
#include <stdio.h>
#include <math.h>
-#include <malloc.h>
+#include <sys/malloc.h>
#include <assert.h>
#include <stdlib.h>
diff -ur bow-20010926/docnames.c bow-ken/docnames.c
--- bow-20010926/docnames.c Tue Oct 5 12:24:08 1999
+++ bow-ken/docnames.c Fri Dec 14 01:13:25 2001
@@ -146,6 +146,21 @@
return num_files;
}
+struct append_filename_context {
+ bow_doc_list *dl_next;
+ bow_doc_list **dl;
+};
+#define CONTEXT ((struct append_filename_context*)context)
+
+ int append_filename (const char *filename, void *context)
+ {
+ CONTEXT->dl_next = *(CONTEXT->dl);
+ *(CONTEXT->dl) = bow_malloc (sizeof (bow_doc_list) + strlen (filename) + 1);
+ (*(CONTEXT->dl))->next = CONTEXT->dl_next;
+ strcpy ((*(CONTEXT->dl))->filename, filename);
+ return 0;
+ }
+
/* Create a linked list of filenames, and append the document list
pointed to by DL to it; return the new concatenated lists in *DL.
The function returns the total number of filenames. When creating
@@ -157,17 +172,10 @@
const char *dirname,
const char *exclude_patterns)
{
- bow_doc_list *dl_next = NULL;
- int append_filename (const char *filename, void *context)
- {
- dl_next = *dl;
- *dl = bow_malloc (sizeof (bow_doc_list) + strlen (filename) + 1);
- (*dl)->next = dl_next;
- strcpy ((*dl)->filename, filename);
- return 0;
- }
-
- return bow_map_filenames_from_dir (append_filename, NULL,
+ struct append_filename_context context;
+ context.dl_next = NULL;
+ context.dl = dl;
+ return bow_map_filenames_from_dir (append_filename, &context,
dirname, exclude_patterns);
}
diff -ur bow-20010926/info_gain.c bow-ken/info_gain.c
--- bow-20010926/info_gain.c Tue Nov 16 15:28:37 1999
+++ bow-ken/info_gain.c Fri Dec 14 01:14:50 2001
@@ -520,16 +520,7 @@
float *wi2ig; /* the array of information gains */
int wi2ig_size;
int wi, i;
- struct wiig { int wi; float ig; } *wiigs;
- int wiig_compare (const void *wiig1, const void *wiig2)
- {
- if (((struct wiig*)wiig1)->ig > ((struct wiig*)wiig2)->ig)
- return -1;
- else if (((struct wiig*)wiig1)->ig == ((struct wiig*)wiig2)->ig)
- return 0;
- else
- return 1;
- }
+ wiig *wiigs;
wi2ig = bow_infogain_per_wi_new (barrel, num_classes, &wi2ig_size);
if (num_to_print == 0)
@@ -537,7 +528,7 @@
/* Create and fill and array of `word-index and information-gain
structures' that can be sorted. */
- wiigs = bow_malloc (wi2ig_size * sizeof (struct wiig));
+ wiigs = bow_malloc (wi2ig_size * sizeof (wiig));
for (wi = 0; wi < wi2ig_size; wi++)
{
wiigs[wi].wi = wi;
@@ -545,7 +536,7 @@
}
/* Sort it. */
- qsort (wiigs, wi2ig_size, sizeof (struct wiig), wiig_compare);
+ qsort (wiigs, wi2ig_size, sizeof (wiig), wiig_compare);
/* Print it. */
for (i = 0; i < num_to_print; i++)
diff -ur bow-20010926/int4word.c bow-ken/int4word.c
--- bow-20010926/int4word.c Sun Jan 21 17:42:32 2001
+++ bow-ken/int4word.c Fri Dec 14 01:21:46 2001
@@ -304,36 +304,22 @@
int wi2ig_size;
bow_int4str *new_map;
int wi;
- struct wiig_list_entry {
- float ig;
- int wi;
- } *wiig_list;
- /* For sorting the above entries. */
- int compare_wiig_list_entry (const void *e1, const void *e2)
- {
- if (((struct wiig_list_entry*)e1)->ig >
- ((struct wiig_list_entry*)e2)->ig)
- return -1;
- else if (((struct wiig_list_entry*)e1)->ig ==
- ((struct wiig_list_entry*)e2)->ig)
- return 0;
- else return 1;
- }
+ wiig *wiig_list;
new_map = bow_int4str_new (0);
wi2ig = bow_infogain_per_wi_new (barrel, num_classes, &wi2ig_size);
/* Make a list of the info gain numbers paired with their WI's,
in prepartion for sorting. */
- wiig_list = alloca (sizeof (struct wiig_list_entry) * wi2ig_size);
+ wiig_list = alloca (sizeof (wiig) * wi2ig_size);
for (wi = 0; wi < wi2ig_size; wi++)
{
wiig_list[wi].wi = wi;
wiig_list[wi].ig = wi2ig[wi];
}
/* Sort the list */
- qsort (wiig_list, wi2ig_size, sizeof (struct wiig_list_entry),
- compare_wiig_list_entry);
+ qsort (wiig_list, wi2ig_size, sizeof (wiig),
+ wiig_compare);
if (num_words_to_keep > wi2ig_size || num_words_to_keep <= 0)
@@ -387,32 +373,45 @@
return total_word_count;
}
-/* Add to the word occurrence counts by recursively decending directory
- DIRNAME and parsing all the text files; skip any files matching
- EXCEPTION_NAME. */
-int
-bow_words_add_occurrences_from_text_dir (const char *dirname,
- const char *exception_name)
-{
- int text_document_count = 0;
- int file_word_count, total_word_count = 0;
+struct words_index_file_context {
+ const char *exception_name;
+ int text_document_count;
+ int file_word_count;
+ int total_word_count;
+};
+#define CONTEXT ((struct words_index_file_context *)context)
+
int words_index_file (const char *filename, void *context)
{
/* If the filename matches the exception name, return immediately. */
- if (exception_name && !strcmp (filename, exception_name))
+ if (CONTEXT->exception_name && !strcmp (filename, CONTEXT->exception_name))
return 0;
- file_word_count = bow_words_add_occurrences_from_file (filename);
- total_word_count += file_word_count;
- if (file_word_count)
- text_document_count++;
+ CONTEXT->file_word_count = bow_words_add_occurrences_from_file (filename);
+ CONTEXT->total_word_count += CONTEXT->file_word_count;
+ if (CONTEXT->file_word_count)
+ CONTEXT->text_document_count++;
return 0;
}
+/* Add to the word occurrence counts by recursively decending directory
+ DIRNAME and parsing all the text files; skip any files matching
+ EXCEPTION_NAME. */
+int
+bow_words_add_occurrences_from_text_dir (const char *dirname,
+ const char *exception_name)
+{
+ struct words_index_file_context context;
+
+ context.text_document_count =
+ context.file_word_count =
+ context.total_word_count = 0;
+ context.exception_name = exception_name;
+
bow_verbosify (bow_progress,
"Counting words... files : unique-words :: "
" ");
- bow_map_filenames_from_dir (words_index_file, 0, dirname, exception_name);
+ bow_map_filenames_from_dir (words_index_file, &context, dirname, exception_name);
bow_verbosify (bow_progress, "\n");
- return total_word_count;
+ return context.total_word_count;
}
diff -ur bow-20010926/maxent.c bow-ken/maxent.c
--- bow-20010926/maxent.c Thu Sep 14 16:47:05 2000
+++ bow-ken/maxent.c Fri Dec 14 02:22:37 2001
@@ -314,25 +314,15 @@
long total_words = 0;
int max_wi = MIN (barrel->wi2dvf->size, bow_num_words());
int max_ci = bow_barrel_num_classes (barrel);
- struct wiig { int wi; float ig; } **mis;
+ wiig **mis;
int wpi;
- int wiig_compare (const void *wiig1, const void *wiig2)
- {
- if (((struct wiig*)wiig1)->ig > ((struct wiig*)wiig2)->ig)
- return -1;
- else if (((struct wiig*)wiig1)->ig == ((struct wiig*)wiig2)->ig)
- return 0;
- else
- return 1;
- }
-
/* do this on a class barrel */
/* malloc and initialize the double array of mutual infos */
mis = bow_malloc (sizeof(struct wiig *) * max_ci);
for (ci = 0; ci < max_ci ; ci ++)
- mis[ci] = bow_malloc (sizeof (struct wiig) * max_wi);
+ mis[ci] = bow_malloc (sizeof (wiig) * max_wi);
for (ci = 0; ci < max_ci ; ci++)
for (wi = 0 ; wi < max_wi; wi++)
@@ -378,7 +368,7 @@
/* ok, now sort each class to bring the best infogains to the top*/
for (ci = 0; ci < max_ci; ci++)
- qsort(mis[ci], max_wi, sizeof (struct wiig), wiig_compare);
+ qsort(mis[ci], max_wi, sizeof (wiig), wiig_compare);
/* Check that we're not saving bogus words */
for (ci = 0; ci < max_ci; ci++)
diff -ur bow-20010926/multiclass.c bow-ken/multiclass.c
--- bow-20010926/multiclass.c Fri Oct 15 15:17:42 1999
+++ bow-ken/multiclass.c Fri Dec 14 02:37:04 2001
@@ -1047,9 +1047,6 @@
}
}
-void
-bow_sort_scores (bow_score *scores, int count)
-{
static int score_compare (const void *x, const void *y)
{
if (((bow_score *)x)->weight > ((bow_score *)y)->weight)
@@ -1059,6 +1056,10 @@
else
return 1;
}
+
+void
+bow_sort_scores (bow_score *scores, int count)
+{
qsort (scores, count, sizeof (bow_score), score_compare);
}
diff -ur bow-20010926/rainbow.c bow-ken/rainbow.c
--- bow-20010926/rainbow.c Thu Sep 14 16:47:06 2000
+++ bow-ken/rainbow.c Fri Dec 14 02:20:56 2001
@@ -1533,31 +1533,22 @@
}
}
-
-
-/* Run test trials, outputing results to TEST_FP. The results are
- indended to be read and processed by the Perl script
- ./rainbow-stats. The test documents come from files inside the
- directories that are named in argv[]. */
-void
-rainbow_test_files (FILE *out_fp, const char *test_dirname)
-{
+struct process_wv_context {
+ FILE *out_fp;
bow_score *hits;
- /* int num_test_docs; */
- int num_hits_to_retrieve = bow_barrel_num_classes (rainbow_class_barrel);
- int actual_num_hits;
- int hi; /* hit index */
const char *current_class;
+ int actual_num_hits;
+ int num_hits_to_retrieve;
int current_ci;
- int ci;
- unsigned int dirlen = 1024;
- char dir[dirlen];
+};
+#define CONTEXT ((struct process_wv_context *)context)
/* Deals with the word vector once it has been taken from the file
or HDB database. Called by test_file and test_hdb_file. (see below) */
int process_wv (const char *filename, bow_wv *query_wv, void *context)
{
bow_cdoc *class_cdoc;
+ int hi;
if (!query_wv)
{
@@ -1565,9 +1556,9 @@
return 0;
}
- fprintf (out_fp, "%s %s ",
+ fprintf (CONTEXT->out_fp, "%s %s ",
filename, /* This test instance */
- current_class); /* The name of the correct class */
+ CONTEXT->current_class); /* The name of the correct class */
/* Remove words not in the class_barrel */
bow_wv_prune_words_not_in_wi2dvf (query_wv,
@@ -1575,33 +1566,33 @@
bow_wv_set_weights (query_wv, rainbow_class_barrel);
bow_wv_normalize_weights (query_wv, rainbow_class_barrel);
- actual_num_hits =
+ CONTEXT->actual_num_hits =
bow_barrel_score (rainbow_class_barrel,
- query_wv, hits,
- num_hits_to_retrieve,
+ query_wv, CONTEXT->hits,
+ CONTEXT->num_hits_to_retrieve,
(rainbow_arg_state.loo_cv
- ? current_ci
+ ? CONTEXT->current_ci
: -1));
- for (hi = 0; hi < actual_num_hits; hi++)
+ for (hi = 0; hi < CONTEXT->actual_num_hits; hi++)
{
class_cdoc =
bow_array_entry_at_index (rainbow_class_barrel->cdocs,
- hits[hi].di);
+ CONTEXT->hits[hi].di);
/* For the sake CommonLisp, don't print numbers smaller than
1e-35, because it can't `(read)' them. */
if (rainbow_arg_state.use_lisp_score_truncation
- && hits[hi].weight < 1e-35
- && hits[hi].weight > 0)
- hits[hi].weight = 0;
- fprintf (out_fp, "%s:%.*g ",
+ && CONTEXT->hits[hi].weight < 1e-35
+ && CONTEXT->hits[hi].weight > 0)
+ CONTEXT->hits[hi].weight = 0;
+ fprintf (CONTEXT->out_fp, "%s:%.*g ",
filename_to_classname (class_cdoc->filename),
- bow_score_print_precision, hits[hi].weight);
+ bow_score_print_precision, CONTEXT->hits[hi].weight);
}
- fprintf (out_fp, "\n");
+ fprintf (CONTEXT->out_fp, "\n");
return 0;
}
- /* This nested function is called once for each test document. */
+ /* This (formerly) nested function is called once for each test document. */
int test_file (const char *filename, void *context)
{
bow_wv *query_wv = NULL;
@@ -1625,6 +1616,20 @@
return process_wv (filename, query_wv, context);
}
+
+
+/* Run test trials, outputing results to TEST_FP. The results are
+ indended to be read and processed by the Perl script
+ ./rainbow-stats. The test documents come from files inside the
+ directories that are named in argv[]. */
+void
+rainbow_test_files (FILE *out_fp, const char *test_dirname)
+{
+ struct process_wv_context context;
+ int ci;
+ unsigned int dirlen = 1024;
+ char dir[dirlen];
+
#if HAVE_HDB
/* This is used for the case that we are dealing with HDB files.
At this point, the fulltext of the file has already been retrieved
@@ -1644,7 +1649,9 @@
}
#endif
- hits = alloca (sizeof (bow_score) * num_hits_to_retrieve);
+ context.num_hits_to_retrieve = bow_barrel_num_classes (rainbow_class_barrel);
+ context.hits = alloca (sizeof (bow_score) * context.num_hits_to_retrieve);
+ context.out_fp = out_fp;
fprintf (out_fp, "#0\n");
for (ci = 0; ci < bow_barrel_num_classes (rainbow_doc_barrel); ci++)
@@ -1660,15 +1667,15 @@
/* Remember which classname this comes from, so, above, we know
the correct class */
- current_ci = class_cdoc->class;
- current_class = bow_barrel_classname_at_index (rainbow_doc_barrel, ci);
+ context.current_ci = class_cdoc->class;
+ context.current_class = bow_barrel_classname_at_index (rainbow_doc_barrel, ci);
/* Test each document in that diretory. */
#if HAVE_HDB
if (bow_hdb)
bow_map_filenames_from_hdb (test_hdb_file, 0, dir, "");
else
#endif
- bow_map_filenames_from_dir (test_file, 0, dir, "");
+ bow_map_filenames_from_dir (test_file, &context, dir, "");
}
}
diff -ur bow-20010926/Makefile.in bow-ken/Makefile.in
--- bow-20010926/Makefile.in Thu Jan 17 02:18:49 2002
+++ bow-ken/Makefile.in Thu Jan 17 02:20:03 2002
@@ -101,7 +101,7 @@
# Pattern rule
ALL_CPPFLAGS = $(CPPFLAGS) $(INCLUDEFLAGS) -Ibow -I$(srcdir) -I$(srcdir)/argp $(DEFS)
ALL_CFLAGS = $(CFLAGS)
-ALL_LIBS = $(LIBS) -L. -lbow -L./argp -largp -lm -lcrypt
+ALL_LIBS = $(LIBS) -L. -lbow -L./argp -largp -lm
# Libbow section
@@ -116,6 +116,7 @@
barrel.c \
bitvec.c \
bmalloc.c \
+compare.c \
deflexer.c \
dv.c \
docnames.c \
diff -ur bow-20010926/compare.c bow-ken/compare.c
--- bow-20010926/compare.c Thu Jan 17 02:27:50 2002
+++ bow-ken/compare.c Thu Jan 17 02:28:22 2002
@@ -0,0 +1,23 @@
+#include <bow/libbow.h>
+
+ int wiig_compare (const void *wiig1, const void *wiig2)
+ {
+ if (((wiig*)wiig1)->ig > ((wiig*)wiig2)->ig)
+ return -1;
+ else if (((wiig*)wiig1)->ig == ((wiig*)wiig2)->ig)
+ return 0;
+ else
+ return 1;
+ }
+
+ /* For sorting the combined list of document hits */
+ int compare_doc_hits (struct _doc_hit *hit1, struct _doc_hit *hit2)
+ {
+ if (hit1->score < hit2->score)
+ return 1;
+ else if (hit1->score == hit2->score)
+ return 0;
+ else
+ return -1;
+ }
+