ead2ba1 · opencog-link-grammar

Changed files (18) +30 -30

configure.ac +3 -3

…	…	@@ -156,7 +156,7 @@ AC_SUBST(HOST_OS)
156	156
157	157	# ====================================================================
158	158	# FreeBSD work-around. Apparently, the AX_PTHREAD autoconf macro
159		# fails to include -lstdthreads in it's results. See bug report
	159	# fails to include -lstdthreads in its results. See bug report
160	160	# https://github.com/opencog/link-grammar/issues/1355
161	161	# So we hack, and explicitly set it here.
162	162
…	…	@@ -495,7 +495,7 @@ AC_SUBST(SQLITE3_CFLAGS)
495	495	# The AtomSpace dictionary backend
496	496	# TODO - use pkg_config for find the libs.
497	497	# TODO - do NOT specify -lpersist-rocks -lpersist-cog here,
498		# they need to be dynmically loaded instead (for proper
	498	# they need to be dynamically loaded instead (for proper
499	499	# initialization of the shared library ctors these contain).
500	500	# Alternately, if they are found, then wrap them with
501	501	# "-Wl,--no-as-needed"
…	…	@@ -930,7 +930,7 @@ fi
930	930
931	931	# ===================================================================
932	932	# swig is needed for compiling the Perl and Python bindings ...
933		# ... well, actually, no, its not. 'make dist' is currently set up to
	933	# ... well, actually, no, it's not. 'make dist' is currently set up to
934	934	# package all of the files generated by swig, so the user does not need
935	935	# to actually install it. However, swig is needed to create the package,
936	936	# and also needed to build from a GitHub pull.

link-grammar/api-structures.h +1 -1

…	…	@@ -24,7 +24,7 @@
24	24	*
25	25	* To make the API simpler, each of these is typedef'ed as a pointer
26	26	* to a data structure. If you're not used to this, some of the code
27		* may look strange, since its not plain that these types are pointers.
	27	* may look strange, since it's not plain that these types are pointers.
28	28	*
29	29	*********************************************************************/
30	30

link-grammar/dict-atomese/lookup-atomese.cc +1 -1

…	…	@@ -644,7 +644,7 @@ void and_enchain_right(Pool_desc* pool, Exp* &andhead, Exp* &andtail, Exp* item)
644	644	static void report_dict_usage(Dictionary dict)
645	645	{
646	646	// We could also print pool_num_elements_issued() but this is not
647		// interesting; its slightly less than pool_size().
	647	// interesting; it's slightly less than pool_size().
648	648	logger().info("LG Dict: %lu entries; %lu Exp_pool elts; %lu MiBytes",
649	649	dict->num_entries,
650	650	pool_size(dict->Exp_pool),

link-grammar/dict-atomese/sections.cc +1 -1

…	…	@@ -351,7 +351,7 @@ Dict_node * lookup_section(Dictionary dict, const Handle& germ)
351	351	Exp* andtail = nullptr;
352	352
353	353	// We really expect dn to not be null here, but ... perhaps it
354		// is, if its just not been observed before.
	354	// is, if it's just not been observed before.
355	355	Exp* eee = nullptr;
356	356	if (dn)
357	357	{

link-grammar/dict-atomese/word-pairs.cc +3 -3

…	…	@@ -260,7 +260,7 @@ static Exp* make_pair_exprs(Dictionary dict, const Handle& germ)
260	260	// Get the cached link-name for this pair.
261	261	const std::string& slnk = cached_linkname(local, rawpr);
262	262
263		// Direction is easy to determine: its either left or right.
	263	// Direction is easy to determine: it's either left or right.
264	264	char cdir = '+';
265	265	if (rawpr->getOutgoingAtom(1) == germ) cdir = '-';
266	266
…	…	@@ -446,7 +446,7 @@ static Exp* make_any_conns(Dictionary dict, Pool_desc* pool)
446	446	/// and arity is 3, then this will return `(A+ or B- or C+ or ())
447	447	/// and (A+ or B- or C+ or ()) and (A+ or B- or C+ or ())`. When
448	448	/// this is exploded into disjuncts, any combination is possible,
449		/// from size zero to three. That's why its a Cartesian product.
	449	/// from size zero to three. That's why it's a Cartesian product.
450	450	///
451	451	/// FYI, this is a work-around for the lack of a commmutative
452	452	/// multi-product. What we really want to do here is to have the
…	…	@@ -497,7 +497,7 @@ Exp* make_cart_pairs(Dictionary dict, const Handle& germ,
497	497
498	498	Exp* optex = make_optional_node(pool, epr);
499	499
500		// If its 1-dimensional, we are done.
	500	// If it's 1-dimensional, we are done.
501	501	if (1 == arity) return optex;
502	502
503	503	Exp* andhead = nullptr;

link-grammar/dict-common/dict-common.h +1 -1

…	…	@@ -151,7 +151,7 @@ struct Dictionary_s
151	151	/* Duplicate words are disallowed in 4.0.dict unless
152	152	* allow_duplicate_words is defined to "true".
153	153	* Duplicate idioms are allowed, unless the "test" parse option
154		* is set to "disalow-dup-idioms" (listing them for debug).
	154	* is set to "disallow-dup-idioms" (listing them for debug).
155	155	* If these variables are 0, they get their allow/disallow values
156	156	* when the first duplicate word/idiom is encountered.
157	157	* 0: not set; 1: allow; -1: disallow */

link-grammar/dict-common/dict-utils.c +1 -1

…	…	@@ -36,7 +36,7 @@ void patch_subscript(char * s)
36	36	if (*de == '\0') return;
37	37	dp = (int) *de;
38	38
39		/* If it's followed by a UTF8 char, its NOT a subscript */
	39	/* If it's followed by a UTF8 char, it's NOT a subscript */
40	40	if (127 < dp \|\| dp < 0) return;
41	41	/* assert ((0 < dp) && (dp <= 127), "Bad dictionary entry!"); */
42	42	if (isdigit(dp)) return;

link-grammar/dict-common/idiom.c +1 -1

…	…	@@ -223,7 +223,7 @@ void insert_idiom(Dictionary dict, Dict_node * dn)
223	223
224	224	/* ---- end of the code alluded to above ---- */
225	225
226		/* now its time to insert them into the dictionary */
	226	/* now it's time to insert them into the dictionary */
227	227
228	228	dn_list = start_dn_list;
229	229

link-grammar/dict-ram/dict-ram.c +2 -2

…	…	@@ -69,7 +69,7 @@ void free_dictionary_root(Dictionary dict)
69	69	/**
70	70	* dict_order_strict - order two dictionary words in proper sort order.
71	71	* Return zero if the strings match, else return in a unique order.
72		* The order is NOT (locale-dependent) UTF8 sort order; its ordered
	72	* The order is NOT (locale-dependent) UTF8 sort order; it's ordered
73	73	* based on numeric values of single bytes. This will uniquely order
74	74	* UTF8 strings, just not in a LANG-dependent (locale-dependent) order.
75	75	* But we don't need/want locale-dependent ordering!
…	…	@@ -309,7 +309,7 @@ Dict_node * dict_node_wild_lookup(Dictionary dict, const char *s)
309	309	* The in-RAM representation is NOT a binary tree; instead it is a tree
310	310	* of lists. An Exp node contains a tag: `AND_type`, `OR_type`, etc.
311	311	* The `operand_next` field is used to hold a linked list which is
312		* joined up with the given Exp type. This is more efficent to traverse,
	312	* joined up with the given Exp type. This is more efficient to traverse,
313	313	* and also saves space, as compared to an ordinary binary tree.
314	314	*
315	315	* Example: (A or B or C or D) becomes

link-grammar/dict-sql/read-sql.c +1 -1

…	…	@@ -66,7 +66,7 @@ static const char * make_expression(Dictionary dict,
66	66	while (p && (lg_isspace((unsigned char)p))) p++;
67	67	if (0 == *p) return p;
68	68
69		/* If it's an open paren, assume its the beginning of a new list */
	69	/* If it's an open paren, assume it's the beginning of a new list */
70	70	if ('(' == *p)
71	71	{
72	72	p = make_expression(dict, ++p, pex);

link-grammar/parse/count.c +2 -2

…	…	@@ -66,7 +66,7 @@ struct Table_tracon_s
66	66	* Each element of these arrays points to a vector, called here word-vector,
67	67	* with a size equal to abs(nearest_word - farthest_word) + 1.
68	68	*
69		* Each element of this vector predicts, in it's status field, whether the
	69	* Each element of this vector predicts, in its status field, whether the
70	70	* expected count is zero for a given null-count (when the word it
71	71	* represents is the end word of the range).
72	72	* This prediction is valid for null-counts up to null_count (or for any
…	…	@@ -164,7 +164,7 @@ static size_t estimate_tracon_entries(Sentence sent)
164	164	}
165	165
166	166	#if HAVE_THREADS_H && !__EMSCRIPTEN__
167		/* Each thread will get it's own version of the `kept_table`.
	167	/* Each thread will get its own version of the `kept_table`.
168	168	* If the program creates zillions of threads, then there will
169	169	* be a mem-leak if this table is not released when each thread
170	170	* exits. This code arranges so that `free_tls_table` is called

link-grammar/parse/extract-links.c +1 -1

…	…	@@ -851,7 +851,7 @@ static void issue_links_for_choice(Linkage lkg, Parse_choice *pc,
851	851	*
852	852	* How it works:
853	853	*
854		* Each linkage has the abstact form of a binary tree, with left and
	854	* Each linkage has the abstract form of a binary tree, with left and
855	855	* right subtrees. The Parse_set is an encoding for all possible
856	856	* trees. Selecting a linkage is then a matter of selecting tree from
857	857	* out of the parse-set.

link-grammar/parse/parse.c +2 -2

…	…	@@ -354,7 +354,7 @@ static int linkage_equiv_p(Linkage lpv, Linkage lnx)
354	354	// names were the same. The connector types might still differ,
355	355	// due to intersection. The multi-connector flag might differ.
356	356	// However, neither of these are likely. It is plausible to skip
357		// this check entirely, its mostly a CPU-time-waster that will
	357	// this check entirely, it's mostly a CPU-time-waster that will
358	358	// never find any differences for the almost any situation.
359	359	for (uint32_t li=0; li<lpv->num_links; li++)
360	360	{
…	…	@@ -460,7 +460,7 @@ static void deduplicate_linkages(Sentence sent, int linkage_limit)
460	460	!sent->overflowed && (sent->num_linkages_found <= linkage_limit)))
461	461	return;
462	462
463		// Deduplicate the valid linkages only; its not worth wasting
	463	// Deduplicate the valid linkages only; it's not worth wasting
464	464	// CPU time on the rest. Sorting guarantees that the valid
465	465	// linkages come first.
466	466	uint32_t nl = sent->num_valid_linkages;

link-grammar/parse/prune.c +1 -1

…	…	@@ -44,7 +44,7 @@
44	44	#define PRx(x) fprintf(stderr, ""#x)
45	45	#define PR(...) true
46	46
47		/* Indicator that this connector cannot be used -- that its "obsolete". */
	47	/* Indicator that this connector cannot be used -- that it's "obsolete". */
48	48	#define BAD_WORD (MAX_SENTENCE+1)
49	49
50	50	typedef uint8_t WordIdx_m; /* Storage representation of word index */

link-grammar/prepare/exprune.c +1 -1

…	…	@@ -185,7 +185,7 @@ static inline bool matches_S(connector_table *ct, int w, condesc_t c)
185	185	* doesn't match anything in the set S.
186	186	*
187	187	* If an OR or AND type expression node has one child, we can replace it
188		* by it's child. This, of course, is not really necessary, except for
	188	* by its child. This, of course, is not really necessary, except for
189	189	* performance.
190	190	*/
191	191

link-grammar/tokenize/tokenize.c +6 -6

…	…	@@ -1444,7 +1444,7 @@ static bool mprefix_split(Sentence sent, Gword unsplit_word, const char word)
1444	1444	mprefix_list = AFCLASS(dict->affix_table, AFDICT_MPRE);
1445	1445	mp_strippable = mprefix_list->length;
1446	1446	if (0 == mp_strippable) return false;
1447		/* The mprefix list is revered-sorted according to prefix length.
	1447	/* The mprefix list is reversed-sorted according to prefix length.
1448	1448	* The code here depends on that. */
1449	1449	mprefix = mprefix_list->string;
1450	1450
…	…	@@ -1539,11 +1539,11 @@ static bool mprefix_split(Sentence sent, Gword unsplit_word, const char word)
1539	1539	}
1540	1540
1541	1541	/* Return true if the word might be capitalized by convention:
1542		* -- if its the first word of a sentence
1543		* -- if its the first word following a colon, a period, a question mark,
	1542	* -- if it's the first word of a sentence
	1543	* -- if it's the first word following a colon, a period, a question mark,
1544	1544	* or any bullet (For example: VII. Ancient Rome)
1545		* -- if its the first word following an ellipsis
1546		* -- if its the first word of a quote
	1545	* -- if it's the first word following an ellipsis
	1546	* -- if it's the first word of a quote
1547	1547	*
1548	1548	* XXX FIXME: These rules are rather English-centric. Someone should
1549	1549	* do something about this someday.
…	…	@@ -2270,7 +2270,7 @@ static void issue_r_stripped(Sentence sent,
2270	2270	if (NULL != r_stripped[1][i])
2271	2271	{
2272	2272	/* We are going to issue a subscripted word which is not a
2273		* substring of it's unsplit_word. For now, the token position
	2273	* substring of its unsplit_word. For now, the token position
2274	2274	* computation code needs an indication for that. */
2275	2275	replabel = strdupa(label);
2276	2276	replabel[0] = REPLACEMENT_MARK[0];

link-grammar/tracon-set.c +1 -1

…	…	@@ -64,7 +64,7 @@ static tid_hash_t hash_connectors(const Connector *c, unsigned int shallow)
64	64	* cache/swap trashing if the table temporary grows very big. However, it
65	65	* had a bug, and it is not clear when to shrink the table - shrinking it
66	66	* unnecessarily can cause an overhead of a table growth. Keep for
67		* possible reimlementation of a similar idea.
	67	* possible reimplementation of a similar idea.
68	68	*/
69	69	static unsigned int find_prime_for(size_t count)
70	70	{

link-parser/lg_readline.c +1 -1

…	…	@@ -49,7 +49,7 @@ static wchar_t * prompt(EditLine *el)
49	49	return wc_prompt;
50	50	}
51	51
52		// lg_readline()is called via a chain of functions:
	52	// lg_readline() is called via a chain of functions:
53	53	// fget_input_string -> get_line -> get_terminal_line -> lg_readline.
54	54	// To avoid changing all of them, this variable is static for now.
55	55	// FIXME: Move the call of find_history_filepath() to lg_readline(), and

opencog-link-grammar commits / ead2ba1

Changed files (18) +30 -30