Add comments and a missing CHECK_FOR_INTERRUPTS in ts_headline.

tglsfdc · tglsfdc · commit 5644d6f909b1 · 2022-11-21T17:07:29.000-05:00
I just spent an annoying amount of time reverse-engineering the 100%-undocumented API between ts_headline and the text search parser's prsheadline function. Add some commentary about that while it's fresh in mind. Also remove some unused macros in wparser_def.c. While at it, I noticed that when commit 78e73e8 added a CHECK_FOR_INTERRUPTS call in TS_execute_recurse, it missed doing so in the parallel function TS_phrase_execute, which surely needs one just as much. Back-patch because of the missing CHECK_FOR_INTERRUPTS. Might as well back-patch the rest of this too.
diff --git a/src/backend/tsearch/ts_parse.c b/src/backend/tsearch/ts_parse.c
@@ -433,6 +433,8 @@ parsetext(Oid cfgId, ParsedText *prs, char *buf, int buflen)
 /*
  * Headline framework
  */
+
+/* Add a word to prs->words[] */
 static void
 hladdword(HeadlineParsedText *prs, char *buf, int buflen, int type)
 {
@@ -449,6 +451,14 @@ hladdword(HeadlineParsedText *prs, char *buf, int buflen, int type)
 	prs->curwords++;
 }
 
+/*
+ * Add pos and matching-query-item data to the just-added word.
+ * Here, buf/buflen represent a processed lexeme, not raw token text.
+ *
+ * If the query contains more than one matching item, we replicate
+ * the last-added word so that each item can be pointed to.  The
+ * duplicate entries are marked with repeated = 1.
+ */
 static void
 hlfinditem(HeadlineParsedText *prs, TSQuery query, int32 pos, char *buf, int buflen)
 {
@@ -589,6 +599,9 @@ hlparsetext(Oid cfgId, HeadlineParsedText *prs, TSQuery query, char *buf, int bu
 	FunctionCall1(&(prsobj->prsend), PointerGetDatum(prsdata));
 }
 
+/*
+ * Generate the headline, as a text object, from HeadlineParsedText.
+ */
 text *
 generateHeadline(HeadlineParsedText *prs)
 {
diff --git a/src/backend/tsearch/wparser_def.c b/src/backend/tsearch/wparser_def.c
@@ -1914,10 +1914,6 @@ prsd_end(PG_FUNCTION_ARGS)
  */
 
 /* token type classification macros */
-#define LEAVETOKEN(x)	( (x)==SPACE )
-#define COMPLEXTOKEN(x) ( (x)==URL_T || (x)==NUMHWORD || (x)==ASCIIHWORD || (x)==HWORD )
-#define ENDPUNCTOKEN(x) ( (x)==SPACE )
-
 #define TS_IDIGNORE(x)	( (x)==TAG_T || (x)==PROTOCOL || (x)==SPACE || (x)==XMLENTITY )
 #define HLIDREPLACE(x)	( (x)==TAG_T )
 #define HLIDSKIP(x)		( (x)==URL_T || (x)==NUMHWORD || (x)==ASCIIHWORD || (x)==HWORD )
diff --git a/src/backend/utils/adt/tsvector_op.c b/src/backend/utils/adt/tsvector_op.c
@@ -1617,6 +1617,9 @@ TS_phrase_execute(QueryItem *curitem, void *arg, uint32 flags,
 	/* since this function recurses, it could be driven to stack overflow */
 	check_stack_depth();
 
+	/* ... and let's check for query cancel while we're at it */
+	CHECK_FOR_INTERRUPTS();
+
 	if (curitem->type == QI_VAL)
 		return chkcond(arg, (QueryOperand *) curitem, data);
 
diff --git a/src/include/tsearch/ts_public.h b/src/include/tsearch/ts_public.h
@@ -30,33 +30,60 @@ typedef struct
 } LexDescr;
 
 /*
- * Interface to headline generator
+ * Interface to headline generator (tsparser's prsheadline function)
+ *
+ * HeadlineParsedText describes the text that is to be highlighted.
+ * Some fields are passed from the core code to the prsheadline function,
+ * while others are output from the prsheadline function.
+ *
+ * The principal data is words[], an array of HeadlineWordEntry,
+ * one entry per token, of length curwords.
+ * The fields of HeadlineWordEntry are:
+ *
+ * in, selected, replace, skip: these flags are initially zero
+ * and may be set by the prsheadline function.  A consecutive group
+ * of tokens marked "in" form a "fragment" to be output.
+ * Such tokens may additionally be marked selected, replace, or skip
+ * to modify how they are shown.  (If you set more than one of those
+ * bits, you get an unspecified one of those behaviors.)
+ *
+ * type, len, pos, word: filled by core code to describe the token.
+ *
+ * item: if the token matches any operand of the tsquery of interest,
+ * a pointer to such an operand.  (If there are multiple matching
+ * operands, we generate extra copies of the HeadlineWordEntry to hold
+ * all the pointers.  The extras are marked with repeated = 1 and should
+ * be ignored except for checking the item pointer.)
  */
 typedef struct
 {
-	uint32		selected:1,
-				in:1,
-				replace:1,
-				repeated:1,
-				skip:1,
-				unused:3,
-				type:8,
-				len:16;
-	WordEntryPos pos;
-	char	   *word;
-	QueryOperand *item;
+	uint32		selected:1,		/* token is to be highlighted */
+				in:1,			/* token is part of headline */
+				replace:1,		/* token is to be replaced with a space */
+				repeated:1,		/* duplicate entry to hold item pointer */
+				skip:1,			/* token is to be skipped (not output) */
+				unused:3,		/* available bits */
+				type:8,			/* parser's token category */
+				len:16;			/* length of token */
+	WordEntryPos pos;			/* position of token */
+	char	   *word;			/* text of token (not null-terminated) */
+	QueryOperand *item;			/* a matching query operand, or NULL if none */
 } HeadlineWordEntry;
 
 typedef struct
 {
+	/* Fields filled by core code before calling prsheadline function: */
 	HeadlineWordEntry *words;
-	int32		lenwords;
-	int32		curwords;
-	int32		vectorpos;		/* positions a-la tsvector */
-	char	   *startsel;
+	int32		lenwords;		/* allocated length of words[] */
+	int32		curwords;		/* current number of valid entries */
+	int32		vectorpos;		/* used by ts_parse.c in filling pos fields */
+
+	/* The prsheadline function must fill these fields: */
+	/* Strings for marking selected tokens and separating fragments: */
+	char	   *startsel;		/* palloc'd strings */
 	char	   *stopsel;
 	char	   *fragdelim;
-	int16		startsellen;
+	int16		startsellen;	/* lengths of strings */
 	int16		stopsellen;
 	int16		fragdelimlen;
 } HeadlineParsedText;
