| 1 |
/*------------------------------------------------------------------------- |
| 2 |
* |
| 3 |
* ts_public.h |
| 4 |
* Public interface to various tsearch modules, such as |
| 5 |
* parsers and dictionaries. |
| 6 |
* |
| 7 |
* Copyright (c) 1998-2025, PostgreSQL Global Development Group |
| 8 |
* |
| 9 |
* src/include/tsearch/ts_public.h |
| 10 |
* |
| 11 |
*------------------------------------------------------------------------- |
| 12 |
*/ |
| 13 |
#ifndef _PG_TS_PUBLIC_H_ |
| 14 |
#define _PG_TS_PUBLIC_H_ |
| 15 |
|
| 16 |
#include "tsearch/ts_type.h" |
| 17 |
|
| 18 |
/* |
| 19 |
* Parser's framework |
| 20 |
*/ |
| 21 |
|
| 22 |
/* |
| 23 |
* returning type for prslextype method of parser |
| 24 |
*/ |
| 25 |
typedef struct |
| 26 |
{ |
| 27 |
int lexid; |
| 28 |
char *alias; |
| 29 |
char *descr; |
| 30 |
} LexDescr; |
| 31 |
|
| 32 |
/* |
| 33 |
* Interface to headline generator (tsparser's prsheadline function) |
| 34 |
* |
| 35 |
* HeadlineParsedText describes the text that is to be highlighted. |
| 36 |
* Some fields are passed from the core code to the prsheadline function, |
| 37 |
* while others are output from the prsheadline function. |
| 38 |
* |
| 39 |
* The principal data is words[], an array of HeadlineWordEntry, |
| 40 |
* one entry per token, of length curwords. |
| 41 |
* The fields of HeadlineWordEntry are: |
| 42 |
* |
| 43 |
* in, selected, replace, skip: these flags are initially zero |
| 44 |
* and may be set by the prsheadline function. A consecutive group |
| 45 |
* of tokens marked "in" form a "fragment" to be output. |
| 46 |
* Such tokens may additionally be marked selected, replace, or skip |
| 47 |
* to modify how they are shown. (If you set more than one of those |
| 48 |
* bits, you get an unspecified one of those behaviors.) |
| 49 |
* |
| 50 |
* type, len, pos, word: filled by core code to describe the token. |
| 51 |
* |
| 52 |
* item: if the token matches any operand of the tsquery of interest, |
| 53 |
* a pointer to such an operand. (If there are multiple matching |
| 54 |
* operands, we generate extra copies of the HeadlineWordEntry to hold |
| 55 |
* all the pointers. The extras are marked with repeated = 1 and should |
| 56 |
* be ignored except for checking the item pointer.) |
| 57 |
*/ |
| 58 |
typedef struct |
| 59 |
{ |
| 60 |
uint32 selected:1, /* token is to be highlighted */ |
| 61 |
in:1, /* token is part of headline */ |
| 62 |
replace:1, /* token is to be replaced with a space */ |
| 63 |
repeated:1, /* duplicate entry to hold item pointer */ |
| 64 |
skip:1, /* token is to be skipped (not output) */ |
| 65 |
unused:3, /* available bits */ |
| 66 |
type:8, /* parser's token category */ |
| 67 |
len:16; /* length of token */ |
| 68 |
WordEntryPos pos; /* position of token */ |
| 69 |
char *word; /* text of token (not null-terminated) */ |
| 70 |
QueryOperand *item; /* a matching query operand, or NULL if none */ |
| 71 |
} HeadlineWordEntry; |
| 72 |
|
| 73 |
typedef struct |
| 74 |
{ |
| 75 |
/* Fields filled by core code before calling prsheadline function: */ |
| 76 |
HeadlineWordEntry *words; |
| 77 |
int32 lenwords; /* allocated length of words[] */ |
| 78 |
int32 curwords; /* current number of valid entries */ |
| 79 |
int32 vectorpos; /* used by ts_parse.c in filling pos fields */ |
| 80 |
|
| 81 |
/* The prsheadline function must fill these fields: */ |
| 82 |
/* Strings for marking selected tokens and separating fragments: */ |
| 83 |
char *startsel; /* palloc'd strings */ |
| 84 |
char *stopsel; |
| 85 |
char *fragdelim; |
| 86 |
int16 startsellen; /* lengths of strings */ |
| 87 |
int16 stopsellen; |
| 88 |
int16 fragdelimlen; |
| 89 |
} HeadlineParsedText; |
| 90 |
|
| 91 |
/* |
| 92 |
* Common useful things for tsearch subsystem |
| 93 |
*/ |
| 94 |
extern char *get_tsearch_config_filename(const char *basename, |
| 95 |
const char *extension); |
| 96 |
|
| 97 |
/* |
| 98 |
* Often useful stopword list management |
| 99 |
*/ |
| 100 |
typedef struct |
| 101 |
{ |
| 102 |
int len; |
| 103 |
char **stop; |
| 104 |
} StopList; |
| 105 |
|
| 106 |
extern void readstoplist(const char *fname, StopList *s, |
| 107 |
char *(*wordop) (const char *, size_t, Oid)); |
| 108 |
extern bool searchstoplist(StopList *s, char *key); |
| 109 |
|
| 110 |
/* |
| 111 |
* Interface with dictionaries |
| 112 |
*/ |
| 113 |
|
| 114 |
/* return struct for any lexize function */ |
| 115 |
typedef struct |
| 116 |
{ |
| 117 |
/*---------- |
| 118 |
* Number of current variant of split word. For example the Norwegian |
| 119 |
* word 'fotballklubber' has two variants to split: ( fotball, klubb ) |
| 120 |
* and ( fot, ball, klubb ). So, dictionary should return: |
| 121 |
* |
| 122 |
* nvariant lexeme |
| 123 |
* 1 fotball |
| 124 |
* 1 klubb |
| 125 |
* 2 fot |
| 126 |
* 2 ball |
| 127 |
* 2 klubb |
| 128 |
* |
| 129 |
* In general, a TSLexeme will be considered to belong to the same split |
| 130 |
* variant as the previous one if they have the same nvariant value. |
| 131 |
* The exact values don't matter, only changes from one lexeme to next. |
| 132 |
*---------- |
| 133 |
*/ |
| 134 |
uint16 nvariant; |
| 135 |
|
| 136 |
uint16 flags; /* See flag bits below */ |
| 137 |
|
| 138 |
char *lexeme; /* C string */ |
| 139 |
} TSLexeme; |
| 140 |
|
| 141 |
/* Flag bits that can appear in TSLexeme.flags */ |
| 142 |
#define TSL_ADDPOS 0x01 |
| 143 |
#define TSL_PREFIX 0x02 |
| 144 |
#define TSL_FILTER 0x04 |
| 145 |
|
| 146 |
/* |
| 147 |
* Struct for supporting complex dictionaries like thesaurus. |
| 148 |
* 4th argument for dictlexize method is a pointer to this |
| 149 |
*/ |
| 150 |
typedef struct |
| 151 |
{ |
| 152 |
bool isend; /* in: marks for lexize_info about text end is |
| 153 |
* reached */ |
| 154 |
bool getnext; /* out: dict wants next lexeme */ |
| 155 |
void *private_state; /* internal dict state between calls with |
| 156 |
* getnext == true */ |
| 157 |
} DictSubState; |
| 158 |
|
| 159 |
#endif /* _PG_TS_PUBLIC_H_ */ |