mirror/ULib
1
0
Fork 0
mirror of https://github.com/stefanocasazza/ULib.git synced 2025-09-28 19:05:55 +08:00
Code Issues Releases Wiki Activity
a308a3387a
ULib/examples/IR/posting.cpp
stefanocasazza 9ac5de60a3 bug fixing
2015-07-13 17:44:12 +02:00

1513 lines
40 KiB
C++
Raw Blame History

// posting.cpp - class support for inverted index (data structure)
#include <ulib/db/rdb.h>
#include <ulib/base/utility.h>
#include <ulib/container/vector.h>
#include <ulib/container/hash_map.h>
#include <ulib/utility/string_ext.h>
#include "posting.h"
extern UCDB* cdb_names;
extern UCDB* cdb_words;
// public
bool UPosting::change_dir;
bool UPosting::ignore_case;
bool UPosting::dir_content_as_doc;
UFile* UPosting::file;
UString* UPosting::word;
UString* UPosting::content;
UString* UPosting::posting;
UString* UPosting::filename;
UString* UPosting::str_cur_doc_id;
uint32_t UPosting::word_freq;
uint32_t UPosting::min_word_size;
uint32_t UPosting::tbl_name_space;
uint32_t UPosting::tbl_words_space;
uint32_t UPosting::max_distance = 2;
UHashMap<UString>* UPosting::tbl_name;
UHashMap<UString>* UPosting::tbl_words;
// private
vPF UPosting::pfunction;
char* UPosting::ptr;
char* UPosting::data;
char* UPosting::ptr_cur_doc_id;
uint64_t UPosting::cur_doc_id;
uint32_t UPosting::pos;
uint32_t UPosting::pos_start;
uint32_t UPosting::space;
uint32_t UPosting::distance;
uint32_t UPosting::size_entry;
uint32_t UPosting::sub_word_size;
uint32_t UPosting::off_last_doc_id;
uint32_t UPosting::vec_sub_word_size;
uint32_t UPosting::sub_word_pos_prev;
uint32_t UPosting::approximate_num_words;
UString* UPosting::sub_word;
UVector<UString>* UPosting::vec_word;
UVector<UString>* UPosting::vec_entry;
UVector<UString>* UPosting::vec_posting;
UVector<UString>* UPosting::vec_sub_word;
UVector<UString>* UPosting::vec_sub_word_posting;
typedef struct u_property {
uint32_t is_meta;
uint32_t is_quoted;
} u_property;
static u_property property[32];
/**
* uint32_t off_last_doc_id;
*
* typedef struct u_posting {
* uint64_t doc_id;
* uint32_t word_freq;
* // -----> array of unsigned with word_freq elements
* } u_posting;
*
* +------+--------------------+--------+-----------+-------+-----+-------+-----+--------+-----------+-------+-----+-------+
* | WORD | offset last DOC id | DOC id | frequency | pos 1 | ... | pos n | ... | DOC id | frequency | pos 1 | ... | pos n |
* +------+--------------------+--------+-----------+-------+-----+-------+-----+--------+-----------+-------+-----+-------+
*/
#define POSTING32(x,attr) u_get_unaligned32(((u_posting*)(x))->attr)
#define POSTING64(x,attr) u_get_unaligned64(((u_posting*)(x))->attr)
#define POSTING_SIZE(x) (sizeof(u_posting)+((POSTING32(x,word_freq))*sizeof(uint32_t)))
#define POSTING_POS(n) u_get_unalignedp32(ptr+sizeof(u_posting)+(n*sizeof(uint32_t)))
#define POSTING_OFFSET_LAST_DOC_ID u_get_unalignedp32(data)
UPosting::UPosting(uint32_t dimension, bool parsing, bool index)
{
U_TRACE_REGISTER_OBJECT(5, UPosting, "%u,%b,%b", dimension, parsing, index)
U_INTERNAL_ASSERT_EQUALS(word,0)
U_INTERNAL_ASSERT_EQUALS(posting,0)
U_INTERNAL_ASSERT_EQUALS(filename,0)
U_INTERNAL_ASSERT_EQUALS(str_cur_doc_id,0)
word = U_NEW(UString);
posting = U_NEW(UString);
filename = U_NEW(UString);
str_cur_doc_id = U_NEW(UString(sizeof(cur_doc_id)));
approximate_num_words = 2000 + (dimension * 8);
if (index)
{
U_INTERNAL_ASSERT_EQUALS(tbl_name, 0)
U_INTERNAL_ASSERT_EQUALS(tbl_words, 0)
dimension += dimension / 4;
tbl_name = U_NEW(UHashMap<UString>(U_GET_NEXT_PRIME_NUMBER(dimension)));
tbl_words = U_NEW(UHashMap<UString>(U_GET_NEXT_PRIME_NUMBER(approximate_num_words), ignore_case));
}
if (parsing)
{
U_INTERNAL_ASSERT_EQUALS(file,0)
U_INTERNAL_ASSERT_EQUALS(content,0)
file = U_NEW(UFile);
content = U_NEW(UString);
}
}
void UPosting::resetVectorCompositeWord()
{
U_TRACE(5, "UPosting::resetVectorCompositeWord()")
delete sub_word;
delete vec_sub_word;
delete vec_sub_word_posting;
sub_word = 0;
vec_sub_word = 0;
vec_sub_word_posting = 0;
vec_sub_word_size = 0;
}
void UPosting::reset()
{
U_TRACE(5, "UPosting::reset()")
max_distance = 2;
sub_word_pos_prev = 0;
if (vec_sub_word) resetVectorCompositeWord();
if (vec_word)
{
delete vec_word;
delete vec_entry;
delete vec_posting;
vec_word = 0;
vec_entry = 0;
vec_posting = 0;
}
}
UPosting::~UPosting()
{
U_TRACE_UNREGISTER_OBJECT(5, UPosting)
delete posting;
delete filename;
delete str_cur_doc_id;
if (tbl_name == 0) reset();
else
{
delete tbl_name;
delete tbl_words;
}
delete word;
if (file)
{
delete file;
delete content;
}
}
// #define U_OPTIMIZE
// #define U_COMPRESS_ENTRY
// MANAGE POSTING VALUE ON DATABASE
inline bool UPosting::decompress()
{
U_TRACE(5, "UPosting::decompress()")
#ifdef U_COMPRESS_ENTRY
if (posting->size() > (sizeof(uint32_t) * 4) &&
UStringExt::isCompress(*posting))
{
posting->decompress();
U_RETURN(true);
}
#endif
U_RETURN(false);
}
U_NO_EXPORT void UPosting::readPosting(UStringRep* word_rep, bool flag)
{
U_TRACE(5, "UPosting::readPosting(%p,%b)", word_rep, flag)
U_INTERNAL_ASSERT_POINTER(cdb_words)
*posting = (word_rep ? (flag ? (*((URDB*)cdb_words))[word_rep]
: (* cdb_words) [word_rep])
: ((URDB*)cdb_words)->UCDB::elem());
#ifdef U_COMPRESS_ENTRY
if (decompress() == false)
#endif
{
// check if add operation...
if (word_rep == 0) posting->duplicate(); // NB: need duplicate string because we need space on string constant..
}
U_INTERNAL_ASSERT_EQUALS(UStringExt::isCompress(posting->data()), false)
}
U_NO_EXPORT int UPosting::writePosting(int flag)
{
U_TRACE(5, "UPosting::writePosting(%d)", flag)
U_INTERNAL_ASSERT_POINTER(cdb_words)
U_INTERNAL_ASSERT_EQUALS(UStringExt::isCompress(posting->data()), false)
#ifdef U_COMPRESS_ENTRY
if (posting->size() > U_CAPACITY) posting->compress();
#endif
int result = ((URDB*)cdb_words)->store(*word, *posting, flag);
U_RETURN(result);
}
// FIND CURRENT DOC ID ON POSTING
inline bool UPosting::checkEntry(char* str, char* s, uint32_t n)
{
U_TRACE(5, "UPosting::checkEntry(%p,%p,%u)", str, n)
U_INTERNAL_ASSERT_POINTER(s)
U_INTERNAL_ASSERT_POINTER(str)
uint32_t offset = (str - s);
U_INTERNAL_DUMP("offset = %u", offset)
U_INTERNAL_DUMP("size_entry = %u", POSTING_SIZE(str))
if ((offset % sizeof(uint32_t)) == 0 &&
POSTING_SIZE(str) <= (n - offset))
{
U_RETURN(true);
}
U_RETURN(false);
}
U_NO_EXPORT __pure char* UPosting::find(char* s, uint32_t n, bool boptmize)
{
U_TRACE(5, "UPosting::find(%p,%u,%b)", s, n, boptmize)
U_INTERNAL_DUMP("cur_doc_id = %llu", cur_doc_id)
if (boptmize)
{
char* str;
do {
str = (char*) u_find(s, n, ptr_cur_doc_id, sizeof(cur_doc_id));
if (str)
{
if (checkEntry(str, s, n)) U_RETURN(str);
++str;
n = (s + n) - str;
s = str;
}
}
while (str);
}
else
{
char* str = s;
char* end = s + n;
do {
if (POSTING64(str,doc_id) == cur_doc_id)
{
U_INTERNAL_DUMP("offset = %u", (str - s))
U_INTERNAL_DUMP("doc_id = %llu", POSTING64(str,doc_id))
U_INTERNAL_DUMP("word_freq = %u", POSTING32(str,word_freq))
U_INTERNAL_DUMP("size_entry = %u", POSTING_SIZE(str))
U_RETURN(str);
}
str += POSTING_SIZE(str);
}
while (str < end);
U_INTERNAL_ASSERT_EQUALS((str - s), (ptrdiff_t)n)
}
U_RETURN((char*)0);
}
U_NO_EXPORT bool UPosting::setPosting(bool bcache)
{
U_TRACE(5, "UPosting::setPosting(%b)", bcache)
U_INTERNAL_ASSERT_POINTER(ptr_cur_doc_id)
if (bcache)
{
static char* last_posting_ptr;
static uint32_t last_posting_sz;
static UStringRep* last_word_rep;
if (word->rep != last_word_rep)
{
readPosting(last_word_rep = word->rep, false);
if (posting->empty()) last_posting_sz = 0;
else
{
last_posting_sz = posting->size() - sizeof(uint32_t); // offset last DOC id
last_posting_ptr = posting->data() + sizeof(uint32_t);
}
}
char* posting_ptr = (last_posting_sz ? find(last_posting_ptr, last_posting_sz, true) : 0);
if (posting_ptr) U_RETURN(true);
U_RETURN(false);
}
U_INTERNAL_ASSERT(*posting)
// if (posting->empty()) U_RETURN(false);
ptr = find(posting->data() + sizeof(uint32_t), posting->size() - sizeof(uint32_t), false);
U_INTERNAL_DUMP("ptr = %p", ptr)
if (ptr) U_RETURN(true);
U_RETURN(false);
}
U_NO_EXPORT bool UPosting::findCurrentDocIdOnPosting(UStringRep* value)
{
U_TRACE(5, "UPosting::findCurrentDocIdOnPosting(%.*S)", U_STRING_TO_TRACE(*value))
posting->_assign(value);
#ifdef U_COMPRESS_ENTRY
(void)decompress();
#endif
ptr = (char*) u_find(U_STRING_TO_PARAM(*posting), ptr_cur_doc_id, sizeof(cur_doc_id));
// check for collision...
if (ptr &&
checkEntry(ptr, U_STRING_TO_PARAM(*posting)))
{
U_RETURN(true);
}
U_RETURN(false);
}
// DELETE WORD
inline bool UPosting::isOneEntry()
{
U_TRACE(5, "UPosting::isOneEntry()")
if (posting->size() == (sizeof(uint32_t) + POSTING_SIZE(ptr))) U_RETURN(true);
U_RETURN(false);
}
U_NO_EXPORT void UPosting::del()
{
U_TRACE(5, "UPosting::del()")
U_INTERNAL_ASSERT(*word)
U_INTERNAL_ASSERT(*posting)
U_INTERNAL_ASSERT_POINTER(ptr)
data = posting->data();
U_INTERNAL_DUMP("word_freq = %u", POSTING32(ptr,word_freq))
U_INTERNAL_DUMP("size_entry = %u", POSTING_SIZE(ptr))
U_INTERNAL_DUMP("cur_doc_id = %llu", cur_doc_id)
U_INTERNAL_DUMP("posting->size() = %u", posting->size())
U_INTERNAL_DUMP("off_last_doc_id = %u", POSTING_OFFSET_LAST_DOC_ID)
U_INTERNAL_ASSERT_MAJOR(POSTING32(ptr,word_freq), 0)
// check if other reference...
int result;
if (isOneEntry())
{
result = ((URDB*)cdb_words)->remove(*word);
if (result != 0)
{
U_ERROR("del(): error<%d> on operation remove word<%.*s> from index database", result, U_STRING_TO_TRACE(*word));
}
return;
}
#if defined(DEBUG) || defined(U_TEST)
uint32_t last_offset = POSTING_OFFSET_LAST_DOC_ID;
#endif
char* end = posting->rep->end();
do {
U_INTERNAL_DUMP("doc_id = %lu", POSTING64(ptr,doc_id))
size_entry = POSTING_SIZE(ptr);
if (POSTING64(ptr,doc_id) == cur_doc_id) break;
ptr += size_entry;
}
while (ptr < end);
if (ptr >= end)
{
U_ERROR("del(): cannot find DocID<%llu> reference for word<%.*s> on index database", cur_doc_id, U_STRING_TO_TRACE(*word));
}
(void) posting->replace(ptr - data, size_entry, 0, '\0');
// reassign value
data = posting->data();
ptr = data + sizeof(uint32_t);
end = posting->rep->end();
U_INTERNAL_DUMP("posting->size() = %u", posting->size())
U_INTERNAL_ASSERT_EQUALS(POSTING_OFFSET_LAST_DOC_ID, last_offset)
do {
size_entry = POSTING_SIZE(ptr);
ptr += size_entry;
}
while (ptr < end);
U_INTERNAL_ASSERT_EQUALS(ptr - data, (ptrdiff_t)posting->size())
u_put_unalignedp32(data, (ptr - data) - size_entry);
U_INTERNAL_DUMP("off_last_doc_id = %u", POSTING_OFFSET_LAST_DOC_ID)
U_INTERNAL_ASSERT_MINOR(POSTING_OFFSET_LAST_DOC_ID, last_offset)
result = writePosting(RDB_REPLACE);
if (result != 0)
{
U_ERROR("del(): error<%d> on operation replace reference for word<%.*s> on index database", result, U_STRING_TO_TRACE(*word));
}
}
// ADD WORD
inline void UPosting::init()
{
U_TRACE(5, "UPosting::init()")
u_put_unalignedp64(&(((u_posting*)(ptr))->doc_id), cur_doc_id);
u_put_unalignedp32(&(((u_posting*)(ptr))->word_freq), 0);
}
inline void UPosting::checkCapacity()
{
U_TRACE(5, "UPosting::checkCapacity()")
space = off_last_doc_id + size_entry + sizeof(off_last_doc_id);
U_INTERNAL_DUMP("space = %u", space)
if (space >= posting->capacity())
{
if (tbl_words) tbl_words_space -= posting->capacity();
// if (posting->uniq() == false) posting->duplicate();
posting->reserve(space * 2);
if (tbl_words)
{
tbl_words_space += posting->capacity();
tbl_words->replaceAfterFind(*posting);
}
}
data = posting->data();
ptr = data + off_last_doc_id;
}
U_NO_EXPORT void UPosting::add()
{
U_TRACE(5, "UPosting::add()")
U_INTERNAL_ASSERT(*word)
bool present = (tbl_words ? tbl_words->find(*word)
: ((URDB*)cdb_words)->find(*word));
if (present == false)
{
// first document for this word...
*posting = UString(U_CAPACITY);
data = posting->data();
ptr = data + sizeof(off_last_doc_id);
// setting off_last_doc_id for first document...
size_entry = sizeof(u_posting);
off_last_doc_id = sizeof(off_last_doc_id);
u_put_unalignedp32(data, sizeof(off_last_doc_id));
init();
if (tbl_words)
{
word->duplicate(); // NB: need duplicate string because depends on mmap()'s content of document...
tbl_words_space += word->size() + posting->capacity();
tbl_words->insertAfterFind(*word, *posting);
}
}
else
{
if (tbl_words) posting->_assign(tbl_words->elem());
else readPosting(0, false);
// U_ASSERT_EQUALS(0,find(posting->data() + sizeof(uint32_t), posting->size() - sizeof(uint32_t)))
data = posting->data();
off_last_doc_id = POSTING_OFFSET_LAST_DOC_ID;
ptr = data + off_last_doc_id;
size_entry = POSTING_SIZE(ptr);
U_INTERNAL_DUMP("word_freq = %u", POSTING32(ptr,word_freq))
U_INTERNAL_DUMP("size_entry = %u", size_entry)
U_INTERNAL_DUMP("off_last_doc_id = %u", off_last_doc_id)
U_INTERNAL_ASSERT_MAJOR(POSTING32(ptr,word_freq), 0)
U_ASSERT((off_last_doc_id + size_entry) <= posting->capacity())
// check for new document...
if (POSTING64(ptr,doc_id) != cur_doc_id)
{
off_last_doc_id += size_entry;
size_entry = sizeof(u_posting);
u_put_unalignedp32(data, off_last_doc_id);
// check string capacity...
checkCapacity();
init();
}
}
U_INTERNAL_ASSERT_EQUALS(cur_doc_id, POSTING64(ptr,doc_id))
U_INTERNAL_ASSERT_EQUALS(off_last_doc_id, POSTING_OFFSET_LAST_DOC_ID)
// check string capacity...
checkCapacity();
U_INTERNAL_ASSERT_EQUALS(cur_doc_id, POSTING64(ptr,doc_id))
U_INTERNAL_ASSERT_EQUALS(off_last_doc_id, POSTING_OFFSET_LAST_DOC_ID)
// save position of word in document...
word_freq = POSTING32(ptr,word_freq);
u_put_unalignedp32(ptr+sizeof(u_posting)+(word_freq*sizeof(uint32_t)), pos);
u_put_unalignedp32(&(((u_posting*)(ptr))->word_freq), word_freq+1);
U_INTERNAL_DUMP("pos = %u", pos)
U_INTERNAL_DUMP("pos[0] = %u", POSTING_POS(0))
U_INTERNAL_DUMP("pos[1] = %u", POSTING_POS(1))
U_INTERNAL_DUMP("word_freq = %u", POSTING32(ptr,word_freq))
posting->size_adjust_force(space);
if (tbl_words == 0)
{
int result = writePosting(present ? RDB_REPLACE : RDB_INSERT);
if (result != 0)
{
U_ERROR("add(): error<%d> on operation store reference for word<%.*s> on index database", result, U_STRING_TO_TRACE(*word));
}
}
}
// CHECK POSTING AND WORD
U_NO_EXPORT void UPosting::checkPosting()
{
U_TRACE(5, "UPosting::checkPosting()")
U_INTERNAL_ASSERT_POINTER(ptr)
U_INTERNAL_DUMP("cur_doc_id = %llu", cur_doc_id)
U_INTERNAL_ASSERT_EQUALS(cur_doc_id, POSTING64(ptr,doc_id))
word_freq = POSTING32(ptr,word_freq);
U_INTERNAL_DUMP("word_freq = %u", word_freq)
U_INTERNAL_DUMP("size_entry = %u", POSTING_SIZE(ptr))
U_INTERNAL_DUMP("offset = %u", posting->distance(ptr))
U_INTERNAL_DUMP("posting->size() = %u", posting->size())
#if defined(U_OPTIMIZE) && defined(DEBUG)
if (cdb_words->ignoreCase() == false)
{
U_INTERNAL_ASSERT_DIFFERS(u_find(content->data(), content->size(), word->data(), word->size()),0)
}
#endif
uint32_t end = content->size(),
sz = word->size();
uint32_t* vpos = (uint32_t*)(ptr+sizeof(u_posting));
char* ptr_word = word->data();
char* ptr_data = content->data();
for (uint32_t i = 0; i < word_freq; ++i)
{
uint32_t start = u_get_unalignedp32(vpos+i);
U_INTERNAL_DUMP("start = %u", start)
if (start >= end ||
u_equal(ptr_word, ptr_data+start, sz, cdb_words->ignoreCase()))
{
U_ERROR("checkPosting(): word<%v> reference at position<%u> for DocID<%llu %v> lost", word->rep, start, cur_doc_id, filename->rep);
}
}
}
// PROCESS WORD
void UPosting::processWord(int32_t op)
{
U_TRACE(5, "UPosting::processWord(%d)", op)
U_INTERNAL_ASSERT_POINTER(content)
if (word->size() < min_word_size) return;
pos = pos_start + (word->data() - content->data());
U_INTERNAL_DUMP("word = %V pos = %u", word->rep, pos)
if (op == 0 ||
op == 1)
{
add(); // add/sub
}
else
{
readPosting(word->rep, (op == 2)); // del
if (posting->empty() ||
setPosting(false) == false)
{
U_ERROR("processWord(%d): word<%v> reference at position<%u> for DocID<%llu %v> lost", op, word->rep, pos, cur_doc_id, filename->rep);
}
if (op == 2) del(); // del
else if (op == 3) checkPosting(); // check
}
}
// CHECK DOCUMENT
U_NO_EXPORT int UPosting::checkDocument(UStringRep* word_rep, UStringRep* value)
{
U_TRACE(5, "UPosting::checkDocument(%.*S,%p)", U_STRING_TO_TRACE(*word_rep), value)
U_INTERNAL_ASSERT(*content)
if (findCurrentDocIdOnPosting(value))
{
word->_assign(word_rep);
U_INTERNAL_DUMP("cdb_words->ignoreCase() = %b", cdb_words->ignoreCase())
# ifndef U_OPTIMIZE
if (cdb_words->ignoreCase() == false &&
u_find(content->data(), content->size(), word->data(), word->size()) == 0)
{
U_RETURN(1);
}
# endif
checkPosting();
}
U_RETURN(1);
}
U_NO_EXPORT int UPosting::checkAllEntry(UStringRep* word_rep, UStringRep* value)
{
U_TRACE(5, "UPosting::checkAllEntry(%.*S,%p)", U_STRING_TO_TRACE(*word_rep), value)
word->_assign(word_rep);
callForPostingAndSetFilename(0);
(void) write(1, U_CONSTANT_TO_PARAM(".")); // CHECK_2
U_RETURN(1);
}
void UPosting::checkAllEntry()
{
U_TRACE(5, "UPosting::checkAllEntry()")
cdb_words->callForAllEntryWithPattern(checkAllEntry, 0);
}
// SUBSTITUTE DOCUMENT
U_NO_EXPORT int UPosting::substitute(UStringRep* word_rep, UStringRep* value)
{
U_TRACE(5, "UPosting::substitute(%.*S,%p)", U_STRING_TO_TRACE(*word_rep), value)
posting->_assign(value);
#ifdef U_COMPRESS_ENTRY
(void) decompress();
#endif
word->_assign(word_rep);
if (setPosting(false)) del();
U_RETURN(1);
}
// PROCESS DOCUMENT
inline void UPosting::setDocID(bool from_inode)
{
U_TRACE(5, "UPosting::setDocID(%b)", from_inode)
if (from_inode) cur_doc_id = - file->inode();
U_INTERNAL_DUMP("cur_doc_id = %llu", cur_doc_id)
str_cur_doc_id->setFromInode(&cur_doc_id);
ptr_cur_doc_id = str_cur_doc_id->data();
}
void UPosting::setDocID(int32_t op)
{
U_TRACE(5, "UPosting::setDocID(%d)", op)
U_DUMP("filename = %.*S", U_STRING_TO_TRACE(*filename))
if (dir_content_as_doc)
{
if (op != 0) // add
{
U_ERROR("setDocID(%d): sorry, not implemented", op);
}
if (change_dir == false) return;
change_dir = false;
*filename = UStringExt::dirname(*filename);
filename->duplicate(); // NB: need duplicate string because depends on volatile buffer of filename...
U_INTERNAL_DUMP("dirname = %V", filename->rep)
}
// insert/fetch/remove into table of docs name
setDocID(true);
if (tbl_name)
{
tbl_name->insert(*str_cur_doc_id, *filename);
tbl_name_space += str_cur_doc_id->size() + filename->size();
}
else
{
int result = 0;
if (op == 0) result = ((URDB*)cdb_names)->store( *str_cur_doc_id, *filename, RDB_INSERT); // add
if (op == 2) result = ((URDB*)cdb_names)->remove(*str_cur_doc_id); // del
if (result != 0)
{
U_ERROR("setDocID(%d): error<%d> for operation<%d> on DocID database", op, result, op);
}
if (op == 1) // sub
{
if (((URDB*)cdb_names)->find(*str_cur_doc_id) == false)
{
U_ERROR("setDocID(%d): cannot find DocID<%llu> on names database", op, cur_doc_id);
}
// SUBSTITUTE: we find all reference for this document in the words database and erase it...
U_INTERNAL_ASSERT_POINTER(cdb_words)
((URDB*)cdb_words)->callForAllEntryWithPattern(substitute, 0);
}
else if (op == 3) // check
{
// CHECK: we find all reference for this document in the words database and check it...
U_INTERNAL_ASSERT_POINTER(cdb_words)
cdb_words->callForAllEntryWithPattern(checkDocument, 0);
}
}
}
// PROCESSING MISC (QUERY)
void UPosting::callForPosting(vPF function)
{
U_TRACE(5, "UPosting::callForPosting(%p)", function)
U_INTERNAL_ASSERT(*posting)
U_INTERNAL_ASSERT_POINTER(function)
data = posting->data();
ptr = data + sizeof(uint32_t);
char* end = posting->rep->end();
do {
word_freq = POSTING32(ptr,word_freq);
size_entry = POSTING_SIZE(ptr);
cur_doc_id = POSTING64(ptr,doc_id);
U_INTERNAL_DUMP("word_freq = %u", word_freq)
U_INTERNAL_DUMP("size_entry = %u", size_entry)
U_INTERNAL_DUMP("cur_doc_id = %llu", cur_doc_id)
function();
ptr += size_entry;
}
while (ptr < end);
U_INTERNAL_ASSERT_EQUALS(ptr - data, (ptrdiff_t)posting->size())
}
U_NO_EXPORT void UPosting::setFilename()
{
U_TRACE(5, "UPosting::setFilename()")
U_INTERNAL_ASSERT(*str_cur_doc_id)
U_INTERNAL_ASSERT_POINTER(cdb_names)
*filename = (*cdb_names)[str_cur_doc_id->rep];
if (filename->empty())
{
U_ERROR("setFilename(): cannot find document name from DocID<%llu> reference for word<%v> on index database", cur_doc_id, word->rep);
}
}
U_NO_EXPORT void UPosting::callForPostingAndSetFilename()
{
U_TRACE(5, "UPosting::callForPostingAndSetFilename()")
str_cur_doc_id->setFromInode(&cur_doc_id);
setFilename();
if (pfunction) pfunction();
}
void UPosting::callForPostingAndSetFilename(vPF function)
{
U_TRACE(5, "UPosting::callForPostingAndSetFilename(%p)", function)
pfunction = function;
callForPosting(callForPostingAndSetFilename);
}
U_NO_EXPORT bool UPosting::setVectorCompositeWord()
{
U_TRACE(5, "UPosting::setVectorCompositeWord()")
U_INTERNAL_ASSERT(*word)
U_INTERNAL_ASSERT_POINTER(cdb_words)
U_INTERNAL_ASSERT_EQUALS(vec_sub_word,0)
U_INTERNAL_ASSERT_EQUALS(vec_sub_word_posting,0)
sub_word = U_NEW(UString);
vec_sub_word = U_NEW(UVector<UString>(*word));
vec_sub_word_posting = U_NEW(UVector<UString>);
vec_sub_word_size = vec_sub_word->size();
for (uint32_t i = 0; i < vec_sub_word_size; ++i)
{
*sub_word = (*vec_sub_word)[i];
U_INTERNAL_DUMP("sub_word = %V", sub_word->rep)
if (sub_word->size() < min_word_size)
{
vec_sub_word_posting->push(UString::getStringNull());
continue;
}
readPosting(sub_word->rep, false);
if (posting->empty())
{
resetVectorCompositeWord();
U_RETURN(false);
}
vec_sub_word_posting->push(*posting);
}
U_RETURN(true);
}
inline bool UPosting::setSubWord(uint32_t i)
{
U_TRACE(5, "UPosting::setSubWord(%u)", i)
U_INTERNAL_ASSERT_POINTER(sub_word)
U_INTERNAL_ASSERT_POINTER(vec_sub_word)
*sub_word = (*vec_sub_word)[i];
sub_word_size = sub_word->size();
U_INTERNAL_DUMP("sub_word = %V", sub_word->rep)
if (sub_word_size < min_word_size)
{
sub_word_pos_prev += 1 + sub_word_size;
U_RETURN(false);
}
U_INTERNAL_DUMP("sub_word_pos_prev = %u", sub_word_pos_prev)
*posting = (*vec_sub_word_posting)[i];
U_INTERNAL_ASSERT(*posting)
U_RETURN(true);
}
inline UString UPosting::extractDocID()
{
U_TRACE(5, "UPosting::extractDocID()")
/**
* +------+--------------------+--------+-----------+-------+-----+-------+-----+--------+-----------+-------+-----+-------+
* | WORD | offset last DOC id | DOC id | frequency | pos 1 | ... | pos n | ... | DOC id | frequency | pos 1 | ... | pos n |
* +------+--------------------+--------+-----------+-------+-----+-------+-----+--------+-----------+-------+-----+-------+
*/
UStringRep* r;
UString s(size_entry);
char* sdata = s.data();
char* sptr = sdata;
for (uint32_t i = 0, n = vec_entry->size(); i < n; ++i)
{
r = vec_entry->UVector<UStringRep*>::at(i);
# ifdef U_COMPRESS_ENTRY
posting->_assign(r);
if (decompress()) r = posting->rep;
# endif
data = r->data();
ptr = data + sizeof(uint32_t);
char* end = r->end();
do {
size_entry = POSTING_SIZE(ptr);
cur_doc_id = POSTING64(ptr,doc_id);
U_INTERNAL_DUMP("word_freq = %u", POSTING32(ptr,word_freq))
U_INTERNAL_DUMP("size_entry = %u", size_entry)
U_INTERNAL_DUMP("cur_doc_id = %llu", cur_doc_id)
U_MEMCPY((void*)sptr, &cur_doc_id, sizeof(cur_doc_id));
ptr += size_entry;
sptr += sizeof(uint64_t);
}
while (ptr < end);
U_INTERNAL_ASSERT_EQUALS(ptr - data, (ptrdiff_t)r->size())
}
s.size_adjust_force(sptr - sdata);
U_RETURN_STRING(s);
}
/**
* ------------------------------------------------
* findDocID() is be called from:
* ------------------------------------------------
* UQueryParser::evaluate() (loop for all doc name)
* ------------------------------------------------
*/
bool UPosting::findDocID(UStringRep* word_rep)
{
U_TRACE(5, "UPosting::findDocID(%V)", word_rep)
U_INTERNAL_ASSERT(*str_cur_doc_id)
U_INTERNAL_ASSERT_POINTER(cdb_words)
// manage meta expr, composite word, etc. for all doc name... (save info with property)
uint32_t i;
word->_assign(word_rep);
if (vec_word) i = vec_word->find(*word);
else
{
// allocation property...
i = U_NOT_FOUND;
vec_word = U_NEW(UVector<UString>(32));
vec_entry = U_NEW(UVector<UString>(approximate_num_words));
vec_posting = U_NEW(UVector<UString>(32));
}
// check if exist property for this word...
if (i == U_NOT_FOUND)
{
// set property for this word...
i = vec_word->size();
vec_word->push(*word);
if (word_rep->isQuoted('"'))
{
property[i].is_meta = false;
property[i].is_quoted = true;
}
else if (word->find_first_of("?*", 0, 2) != U_NOT_FOUND)
{
property[i].is_meta = true;
property[i].is_quoted = false;
}
else
{
property[i].is_meta = property[i].is_quoted = false;
}
}
// find context with property...
if (property[i].is_quoted)
{
// context composite word...
static UStringRep* last_word_rep;
if (word_rep != last_word_rep)
{
last_word_rep = word_rep;
if (vec_sub_word) resetVectorCompositeWord();
}
if (callForCompositeWord(0)) U_RETURN(true);
U_RETURN(false);
}
if (property[i].is_meta)
{
// context meta word...
UString entry;
if (i < vec_posting->size()) entry = (*vec_posting)[i];
else
{
vec_entry->clear();
if (cdb_words->getValuesWithKeyNask(*vec_entry, *word, &size_entry)) entry = extractDocID();
vec_posting->push(entry);
}
if (entry &&
u_find(entry.data(), entry.size(), ptr_cur_doc_id, sizeof(cur_doc_id)))
{
U_RETURN(true);
}
U_RETURN(false);
}
// context single word...
if (setPosting(true)) U_RETURN(true);
U_RETURN(false);
}
void UPosting::callForPosting(vPF function, bool is_space)
{
U_TRACE(5, "UPosting::callForPosting(%p,%b)", function, is_space)
U_INTERNAL_ASSERT(*word)
U_INTERNAL_ASSERT_POINTER(cdb_words)
if (is_space) // composite word
{
// check for NEAR operation (string not quoted)...
bool is_quoted = word->isQuoted();
U_INTERNAL_DUMP("is_quoted = %b", is_quoted)
if (is_quoted == false) max_distance = ((uint32_t)-1);
(void) callForCompositeWord(function);
}
else // single word
{
readPosting(word->rep, false);
if (*posting) callForPostingAndSetFilename(function);
}
}
/**
* ------------------------------------------------
* CONTEXT COMPOSITE WORD...
* ------------------------------------------------
* checkCompositeWord() may be called from:
* ------------------------------------------------
* UQueryParser::evaluate() (loop for all doc name)
* callForPosting() (for composite word)
* ------------------------------------------------
*/
U_NO_EXPORT bool UPosting::callForCompositeWord(vPF function)
{
U_TRACE(5, "UPosting::callForCompositeWord(%p)", function)
U_INTERNAL_ASSERT_POINTER(cdb_names)
if (vec_sub_word == 0 &&
setVectorCompositeWord() == false)
{
U_RETURN(false);
}
// find the first sub-word util
uint32_t first_subword_index;
for (first_subword_index = 0; first_subword_index < vec_sub_word_size; ++first_subword_index)
{
if (setSubWord(first_subword_index) == false) continue;
break;
}
U_INTERNAL_DUMP("FIRST subword util = %.*S", U_STRING_TO_TRACE(*sub_word))
U_INTERNAL_DUMP("first_subword_index = %u", first_subword_index)
U_INTERNAL_ASSERT(*posting)
#if defined(DEBUG) || defined(U_TEST)
uint32_t posting_size = posting->size();
#endif
bool match;
char* ptr1;
char* end1 = 0;
int32_t i, first_subword_freq;
uint32_t j, k, sz = sub_word_size;
// loop for all posting entry of the first sub-word util
// ---------------------------------------------------------------------------------------------------------------
// immatricolazione (-261845,1,1148)(-261811,1,2024)(-261819,4,2037,2624,4075,4567)(-261832,4,2037,2624,4479,4971)
// ---------------------------------------------------------------------------------------------------------------
if (function == 0)
{
if (setPosting(false) == false) U_RETURN(false);
ptr1 = ptr;
goto start;
}
data = posting->data();
ptr1 = data + sizeof(uint32_t);
end1 = posting->rep->end();
do {
ptr = ptr1;
cur_doc_id = POSTING64(ptr,doc_id);
size_entry = POSTING_SIZE(ptr);
U_INTERNAL_DUMP("cur_doc_id = %llu", cur_doc_id)
U_INTERNAL_DUMP("size_entry = %u", size_entry)
setDocID(false);
start:
first_subword_freq = POSTING32(ptr,word_freq);
U_INTERNAL_DUMP("first_subword_freq = %u", first_subword_freq)
U_INTERNAL_ASSERT_MAJOR(first_subword_freq, 0)
i = -1;
match = false;
// loop for all the position of first sub-word util for this document...
loop1:
while (++i < first_subword_freq)
{
ptr = ptr1;
pos = POSTING_POS(i);
sub_word_pos_prev = pos + sz;
U_INTERNAL_DUMP("pos[%3u] = %u", i, pos)
U_INTERNAL_DUMP("sub_word_pos_prev = %u", sub_word_pos_prev)
// check if the other sub-word match in the current doc...
for (j = first_subword_index + 1; j < vec_sub_word_size; ++j)
{
if (setSubWord(j) == false) continue;
if (setPosting(false) == false) goto loop1;
U_INTERNAL_DUMP("subword = %.*S", U_STRING_TO_TRACE(*sub_word))
U_INTERNAL_ASSERT_EQUALS(cur_doc_id, POSTING64(ptr,doc_id))
word_freq = POSTING32(ptr,word_freq);
U_INTERNAL_DUMP("word_freq = %u", word_freq)
U_INTERNAL_ASSERT_MAJOR(word_freq, 0)
// check for all position for this sub-word in the current doc...
for (k = 0; k < word_freq; ++k)
{
pos = POSTING_POS(k);
distance = pos - sub_word_pos_prev;
U_INTERNAL_DUMP("pos[%3u] = %u", k, pos)
U_INTERNAL_DUMP("sub_word_pos_prev = %u", sub_word_pos_prev)
U_INTERNAL_DUMP("distance = %u", distance)
U_INTERNAL_DUMP("max_distance = %u", max_distance)
if (distance <= max_distance) break;
}
if (k == word_freq) goto loop1; // NOT match, go to another position for the first sub-word util...
sub_word_pos_prev = pos + sub_word_size;
}
match = true;
}
U_INTERNAL_ASSERT_EQUALS(i, first_subword_freq)
if (function == 0) U_RETURN(match);
if (match)
{
setFilename();
word_freq = (max_distance == ((uint32_t)-1) // check for NEAR...
? 0
: max_distance - distance);
function();
}
ptr1 += size_entry;
}
while (ptr1 < end1);
U_INTERNAL_ASSERT_EQUALS(ptr1 - data, (ptrdiff_t)posting_size)
U_RETURN(true);
}
// PRINT DATABASE
static ostream* os;
static UString* buffer;
# define SIZE_ENTRY ( 2 + \
10 + \
10 + \
10 * word_freq)
U_NO_EXPORT void UPosting::printPosting()
{
U_TRACE(5, "UPosting::printPosting()")
static UStringRep* last = word->rep;
start:
if (last == word->rep)
{
if (buffer->empty())
{
# ifdef U_STDCPP_ENABLE
*os << "-------------------------------------------------------------------------------------------\n";
last->write(*os);
# endif
}
uint32_t* vpos = (uint32_t*)(ptr+sizeof(u_posting));
if (buffer->space() < SIZE_ENTRY) buffer->reserve(buffer->capacity() * 2);
buffer->snprintf_add("(%llX,%u,%u", cur_doc_id, word_freq, u_get_unalignedp32(vpos));
for (uint32_t i = 1; i < word_freq; ++i) buffer->snprintf_add(",%u", u_get_unalignedp32(vpos+i));
(void) buffer->append(U_CONSTANT_TO_PARAM(")"));
}
else
{
# ifdef U_STDCPP_ENABLE
*os << ' ' << *buffer << '\n';
# endif
last = word->rep;
buffer->setEmpty();
goto start;
}
}
U_NO_EXPORT int UPosting::print(UStringRep* word_rep, UStringRep* value)
{
U_TRACE(5, "UPosting::print(%.*S,%p)", U_STRING_TO_TRACE(*word_rep), value)
posting->_assign(value);
#ifdef U_COMPRESS_ENTRY
(void) decompress();
#endif
word->_assign(word_rep);
callForPosting(printPosting);
U_RETURN(1);
}
U_NO_EXPORT int UPosting::printDocName(UStringRep* doc_id, UStringRep* doc_name)
{
U_TRACE(5, "UPosting::printDocName(%.*S,%.*S)", U_STRING_TO_TRACE(*doc_id), U_STRING_TO_TRACE(*doc_name))
#ifdef U_STDCPP_ENABLE
char _buffer[20];
os->write(_buffer, u__snprintf(_buffer, sizeof(_buffer), "%llX ", *((uint64_t*)(doc_id->data()))));
doc_name->write(*os);
os->put('\n');
#endif
U_RETURN(1);
}
void UPosting::printDB(ostream& s)
{
U_TRACE(5, "UPosting::printDB(%p)", &s)
#ifdef U_STDCPP_ENABLE
os = &s;
buffer = U_NEW(UString(U_CAPACITY));
if (tbl_words) tbl_words->callForAllEntry((bPFprpv)print);
else
{
U_INTERNAL_ASSERT_POINTER(cdb_names)
U_INTERNAL_ASSERT_POINTER(cdb_words)
cdb_names->callForAllEntryWithPattern(printDocName, 0);
cdb_words->callForAllEntryWithPattern(print, 0);
}
*os << ' ' << *buffer
<< "\n-------------------------------------------------------------------------------------------\n";
os->flush();
delete buffer;
#endif
}
// DEBUG
#if defined(DEBUG) && defined(U_STDCPP_ENABLE)
const char* UPosting::dump(bool _reset) const
{
*UObjectIO::os << "pos " << pos << '\n'
<< "ptr " << (void*)ptr << '\n'
<< "data " << (void*)data << '\n'
<< "space " << space << '\n'
<< "distance " << distance << '\n'
<< "pfunction " << (void*)pfunction << '\n'
<< "pos_start " << pos_start << '\n'
<< "size_entry " << size_entry << '\n'
<< "word_freq " << word_freq << '\n'
<< "cur_doc_id " << cur_doc_id << '\n'
<< "max_distance " << max_distance << '\n'
<< "min_word_size " << min_word_size << '\n'
<< "sub_word_size " << sub_word_size << '\n'
<< "tbl_name_space " << tbl_name_space << '\n'
<< "ptr_cur_doc_id " << (void*)ptr_cur_doc_id << '\n'
<< "tbl_words_space " << tbl_words_space << '\n'
<< "off_last_doc_id " << off_last_doc_id << '\n'
<< "vec_sub_word_size " << vec_sub_word_size << '\n'
<< "sub_word_pos_prev " << sub_word_pos_prev << '\n'
<< "file (UFile " << (void*)file << ")\n"
<< "word (UString " << (void*)word << ")\n"
<< "content (UString " << (void*)content << ")\n"
<< "posting (UString " << (void*)posting << ")\n"
<< "filename (UString " << (void*)filename << ")\n"
<< "sub_word (UString " << (void*)sub_word << ")\n"
<< "tbl_name (UHashMap<UString> " << (void*)tbl_name << ")\n"
<< "tbl_words (UHashMap<UString> " << (void*)tbl_words << ")\n"
<< "vec_sub_word (UVector<UString> " << (void*)vec_sub_word << ")\n"
<< "str_cur_doc_id (UString " << (void*)str_cur_doc_id << ")\n"
<< "vec_sub_word_posting (UVector<UString> " << (void*)vec_sub_word_posting << ')';
if (_reset)
{
UObjectIO::output();
return UObjectIO::buffer_output;
}
return 0;
}
#endif
Reference in New Issue View Git Blame Copy Permalink