/*
- * Sherlock Gatherer: Data structures used in indices
+ * Sherlock: Data structures used in indices
*
- * (c) 2001 Martin Mares <mj@ucw.cz>
+ * (c) 2001--2003 Martin Mares <mj@ucw.cz>
*/
-/* Words */
-
-#define MAX_WORD_LEN 64
-
-/* Word types */
-
-enum word_type {
- WT_RESERVED, /* Reserved word type */
- WT_TEXT, /* Ordinary text */
- WT_EMPH, /* Emphasized text */
- WT_SMALL, /* Small font */
- WT_TITLE, /* Document title */
- WT_SMALL_HEADING, /* Heading */
- WT_BIG_HEADING, /* Larger heading */
- WT_KEYWORD, /* Explicitly marked keyword */
- WT_META, /* Various meta-information */
- WT_ALT, /* Alternate texts for graphical elements */
- WT_URL1, /* Word extracted from document URL (low and high weight) */
- WT_URL2,
- WT_MAX
-};
+#ifndef _SHERLOCK_INDEX_H
+#define _SHERLOCK_INDEX_H
-/* Descriptive names used for user output */
-#define WORD_TYPE_USER_NAMES \
- "reserved", "text", "emph", "small", "title", "hdr1", "hdr2", "keywd", \
- "meta", "alt", "urlword1", "urlword2", "type12", "type13", "type14", "type15"
-
-/* Keywords for word type names */
-#define WORD_TYPE_NAMES \
- T(WORD, ~0) \
- T(TEXT, 1 << WT_TEXT) \
- T(EMPH, 1 << WT_EMPH) \
- T(SMALL, 1 << WT_SMALL) \
- T(TITLE, 1 << WT_TITLE) \
- T(HDR, (1 << WT_SMALL_HEADING) | (1 << WT_BIG_HEADING)) \
- T(HDR1, 1 << WT_SMALL_HEADING) \
- T(HDR2, 1 << WT_BIG_HEADING) \
- T(KEYWD, 1 << WT_KEYWORD) \
- T(META, 1 << WT_META) \
- T(ALT, 1 << WT_ALT) \
- T(URLWORD, (1 << WT_URL1) | (1 << WT_URL2))
-
-/* These types are not shown in document contexts */
-#define WORD_TYPES_HIDDEN ((1 << WT_URL1) | (1 << WT_URL2))
-
-/* These types are always matched without accents if accent mode is set to "auto" */
-#define WORD_TYPES_NO_AUTO_ACCENT ((1 << WT_URL1) | (1 << WT_URL2))
-
-/* String types */
-
-enum string_type {
- ST_RESERVED, /* Reserved string type */
- ST_URL, /* URL of the document */
- ST_HOST, /* Host name */
- ST_DOMAIN, /* Domain name */
- ST_REF, /* URL reference */
- ST_BACKREF, /* Back-reference (frame or redirect source) */
- ST_MAX
-};
+#include "lib/fastbuf.h"
+#include SHERLOCK_CUSTOM
+#include "charset/unistream.h"
+
+/*
+ * Words and word complexes
+ *
+ * MAX_WORD_LEN is the maximum length (measured in UTF-8 characters, excluding
+ * the terminating zero byte if there's any) of any word which may appear in the
+ * indices or in the bucket file. Naturally, the same constant also bounds
+ * the number of UCS-2 characters in a word.
+ *
+ * Caveat: If you are upcasing/downcasing the word, the UTF-8 encoding can
+ * expand, although at most twice, so you need to reserve 2*MAX_WORD_LEN bytes.
+ *
+ * MAX_COMPLEX_LEN is the upper bound on number of words in any word complex.
+ */
+
+#define MAX_WORD_LEN 64 /* a multiple of 4 */
+#define MAX_COMPLEX_LEN 10
-#define STRING_TYPE_USER_NAMES \
- "URL", "host", "domain", "ref", "backref", "type5", "type6", "type7", \
- "type8", "type9", "type10", "type11", "type12", "type13", "type14", "type15"
+/* Word and string types are defined in lib/custom.h */
-#define STRING_TYPE_NAMES \
- T(URL, 1 << ST_URL) \
- T(HOST, 1 << ST_HOST) \
- T(DOMAIN, 1 << ST_DOMAIN) \
- T(REF, 1 << ST_REF) \
- T(BACKREF, 1 << ST_BACKREF)
+/* Global index parameters */
-#define STRING_TYPES_URL ((1 << ST_URL) | (1 << ST_REF) | (1 << ST_BACKREF))
-/* These must be indexed in lowercase form */
-#define STRING_TYPES_CASE_INSENSITIVE ((1 << ST_HOST) | (1 << ST_DOMAIN))
+struct index_params {
+ sh_time_t ref_time; /* Reference time (for document ages etc.) */
+};
/* Index card attributes */
struct card_attr {
u32 card; /* Reference to card description (either oid or filepos) */
+#ifdef CONFIG_SITES
u32 site_id;
+#endif
+ CUSTOM_CARD_ATTRS /* Include all custom attributes */
byte weight;
byte flags;
- byte rfu[2];
+#ifdef CONFIG_LASTMOD
+ byte age; /* Document age in pseudo-logarithmic units wrt. reference time */
+#endif
+#ifdef CONFIG_FILETYPE
+ byte type_flags; /* File type flags (see below) */
+#endif
};
enum card_flag {
CARD_FLAG_ACCENTED = 2, /* Document contains accented characters [scanner] */
CARD_FLAG_DUP = 4, /* Removed as a duplicate [merger] */
CARD_FLAG_MERGED = 8, /* Destination of a merge [merger] */
+ CARD_FLAG_IMAGE = 16, /* Is an image object [scanner] */
+ CARD_FLAG_FRAMESET = 32, /* Contains a frameset [scanner] */
};
-#define CARD_POS_SHIFT 5 /* Card positions are shifted this # of bytes to the right */
+#define CARD_POS_SHIFT 5 /* Card positions are shifted this # of bits to the right */
+
+/*
+ * We store document type and several other properties in card_attr->type_flags.
+ * Here we define only the basic structure, the details and also how to match the
+ * types are defined in custom.h.
+ *
+ * bits 7--5 file type: (0-3: text types, 4-7: other types, defined by custom.h)
+ * bits 4--0 type-dependent information, for text types it's document language code
+ */
+
+#define CA_GET_FILE_TYPE(a) ((a)->type_flags >> 5)
+#define CA_GET_FILE_INFO(a) ((a)->type_flags & 0x1f)
+#define CA_GET_FILE_LANG(a) ((a)->type_flags & 0x80 ? 0 : CA_GET_FILE_INFO(a))
/* String fingerprints */
#define GET_TAGGED_CHAR(p,u) do { \
u = *p; \
if (u >= 0xc0) \
- GET_UTF8(p,u); \
+ GET_UTF8_CHAR(p,u); \
else if (u >= 0x80) \
{ \
p++; \
if (u >= 0xb0) \
{ \
- if (u != 0xb0) \
- ASSERT(0); \
+ ASSERT(u == 0xb0); \
u += 0x80020000; \
} \
else if (u >= 0xa0) \
else \
p++; \
} while (0)
+
+#define SKIP_TAGGED_CHAR(p) do { \
+ if (*p >= 0x80 && *p < 0xc0) \
+ { \
+ uns u = *p++; \
+ if (u >= 0xa0 && u < 0xb0 && *p >= 0x80 && *p < 0xc0) \
+ p++; \
+ } \
+ else \
+ UTF8_SKIP(p); \
+} while (0)
+
+static inline uns
+bget_tagged_char(struct fastbuf *f)
+{
+ uns u = bgetc(f);
+ if ((int)u < 0x80)
+ ;
+ else if (u < 0xc0)
+ {
+ if (u >= 0xb0)
+ {
+ ASSERT(u == 0xb0);
+ u += 0x80020000;
+ }
+ else if (u >= 0xa0)
+ {
+ uns v = bgetc(f);
+ ASSERT(v >= 0x80 && v <= 0xbf);
+ u = 0x80010000 + ((u & 0x0f) << 6) + (v & 0x3f);
+ }
+ else
+ u += 0x80000000;
+ }
+ else
+ {
+ bungetc(f);
+ u = bget_utf8(f);
+ }
+ return u;
+}
+
+/* Conversion of document age from seconds to our internal units */
+
+static inline int
+convert_age(sh_time_t lastmod, sh_time_t reftime)
+{
+ sh_time_t age;
+ if (reftime < lastmod) /* past times */
+ return -1;
+ age = (reftime - lastmod) / 3600;
+ if (age < 48) /* last 2 days: 1 hour resolution */
+ return age;
+ age = (age-48) / 24;
+ if (age < 64) /* next 64 days: 1 day resolution */
+ return 48 + age;
+ age = (age-64) / 7;
+ if (age < 135) /* next 135 weeks: 1 week resolution */
+ return 112 + age;
+ age = (age-135) / 52;
+ if (age < 8) /* next 8 years: 1 year resolution */
+ return 247 + age;
+ return 255; /* then just "infinite future" */
+}
+
+#endif