logic: improved fulltext search performance (issue #680)
* move to multithreaded implementation * omit unnecessary work
This commit is contained in:
@@ -29,16 +29,21 @@ std::vector<FullTextSearchResult> FullTextSearchIndex::searchForTerm(const std::
|
||||
TRACE();
|
||||
|
||||
std::vector<FullTextSearchResult> ret;
|
||||
FullTextSearchResult hit;
|
||||
{
|
||||
std::lock_guard<std::mutex> lock(m_filesMutex);
|
||||
for (auto& f : m_files)
|
||||
{
|
||||
FullTextSearchResult hit;
|
||||
hit.fileId = f.fileId;
|
||||
hit.positions = f.array.searchForTerm(term);
|
||||
ret.push_back(hit);
|
||||
std::sort(hit.positions.begin(), hit.positions.end());
|
||||
if (!hit.positions.empty())
|
||||
{
|
||||
ret.push_back(hit);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return ret;
|
||||
}
|
||||
|
||||
|
||||
@@ -552,58 +552,75 @@ std::shared_ptr<SourceLocationCollection> PersistentStorage::getFullTextSearchLo
|
||||
false, true
|
||||
).dispatch();
|
||||
|
||||
const int termLength = searchTerm.length();
|
||||
|
||||
for (const FullTextSearchResult& fileHits : m_fullTextSearchIndex.searchForTerm(searchTerm))
|
||||
{
|
||||
const FilePath filePath = getFileNodePath(fileHits.fileId);
|
||||
std::shared_ptr<TextAccess> fileContent = getFileContent(filePath, false);
|
||||
|
||||
int charsTotal = 0;
|
||||
int lineNumber = 1;
|
||||
std::wstring line = codec.decode(fileContent->getLine(lineNumber));
|
||||
|
||||
for (int pos : fileHits.positions)
|
||||
std::vector<std::shared_ptr<std::thread>> threads;
|
||||
std::mutex collectionMutex;
|
||||
for (std::vector<FullTextSearchResult> fileResults : utility::splitToEqualySizedParts(m_fullTextSearchIndex.searchForTerm(searchTerm), utility::getIdealThreadCount()))
|
||||
{
|
||||
while (charsTotal + (int)line.length() <= pos)
|
||||
{
|
||||
charsTotal += line.length();
|
||||
lineNumber++;
|
||||
line = codec.decode(fileContent->getLine(lineNumber));
|
||||
}
|
||||
std::shared_ptr<std::thread> thread = std::make_shared<std::thread>(
|
||||
[this, &searchTerm, &caseSensitive, &codec, /*no ref here!*/fileResults, &collection, &collectionMutex]()
|
||||
{
|
||||
const int termLength = searchTerm.length();
|
||||
for (const FullTextSearchResult& fileResult : fileResults)
|
||||
{
|
||||
const FilePath filePath = getFileNodePath(fileResult.fileId);
|
||||
std::shared_ptr<TextAccess> fileContent = getFileContent(filePath, false);
|
||||
|
||||
ParseLocation location;
|
||||
location.startLineNumber = lineNumber;
|
||||
location.startColumnNumber = pos - charsTotal + 1;
|
||||
int charsTotal = 0;
|
||||
int lineNumber = 1;
|
||||
std::wstring line = codec.decode(fileContent->getLine(lineNumber));
|
||||
|
||||
if (caseSensitive && line.substr(location.startColumnNumber - 1, termLength) != searchTerm)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
for (int pos : fileResult.positions)
|
||||
{
|
||||
while (charsTotal + (int)line.length() <= pos)
|
||||
{
|
||||
charsTotal += line.length();
|
||||
lineNumber++;
|
||||
line = codec.decode(fileContent->getLine(lineNumber));
|
||||
}
|
||||
|
||||
while ((charsTotal + (int)line.length()) < pos + termLength)
|
||||
{
|
||||
charsTotal += line.length();
|
||||
lineNumber++;
|
||||
line = codec.decode(fileContent->getLine(lineNumber));
|
||||
}
|
||||
ParseLocation location;
|
||||
location.startLineNumber = lineNumber;
|
||||
location.startColumnNumber = pos - charsTotal + 1;
|
||||
|
||||
location.endLineNumber = lineNumber;
|
||||
location.endColumnNumber = pos + termLength - charsTotal;
|
||||
if (caseSensitive && line.substr(location.startColumnNumber - 1, termLength) != searchTerm)
|
||||
{
|
||||
continue;
|
||||
}
|
||||
while ((charsTotal + (int)line.length()) < pos + termLength)
|
||||
{
|
||||
charsTotal += line.length();
|
||||
lineNumber++;
|
||||
line = codec.decode(fileContent->getLine(lineNumber));
|
||||
}
|
||||
location.endLineNumber = lineNumber;
|
||||
location.endColumnNumber = pos + termLength - charsTotal;
|
||||
|
||||
// Set first bit to 1 to avoid collisions
|
||||
const Id locationId = ~(~Id(0) >> 1) + collection->getSourceLocationCount() + 1;
|
||||
|
||||
collection->addSourceLocation(
|
||||
LOCATION_FULLTEXT_SEARCH,
|
||||
locationId,
|
||||
std::vector<Id>(),
|
||||
filePath,
|
||||
location.startLineNumber,
|
||||
location.startColumnNumber,
|
||||
location.endLineNumber,
|
||||
location.endColumnNumber
|
||||
{
|
||||
std::lock_guard<std::mutex> lock(collectionMutex);
|
||||
// Set first bit to 1 to avoid collisions
|
||||
const Id locationId = ~(~Id(0) >> 1) + collection->getSourceLocationCount() + 1;
|
||||
collection->addSourceLocation(
|
||||
LOCATION_FULLTEXT_SEARCH,
|
||||
locationId,
|
||||
std::vector<Id>(),
|
||||
filePath,
|
||||
location.startLineNumber,
|
||||
location.startColumnNumber,
|
||||
location.endLineNumber,
|
||||
location.endColumnNumber
|
||||
);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
);
|
||||
threads.push_back(thread);
|
||||
}
|
||||
|
||||
for (std::shared_ptr<std::thread> thread : threads)
|
||||
{
|
||||
thread->join();
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -5,6 +5,9 @@
|
||||
TextCodec::TextCodec(const std::string& name)
|
||||
: m_name(name)
|
||||
{
|
||||
m_codec = QTextCodec::codecForName(m_name.c_str());
|
||||
m_decoder = std::make_shared<QTextDecoder>(m_codec);
|
||||
m_encoder = std::make_shared<QTextEncoder>(m_codec);
|
||||
}
|
||||
|
||||
std::string TextCodec::getName() const
|
||||
@@ -14,8 +17,7 @@ std::string TextCodec::getName() const
|
||||
|
||||
bool TextCodec::isValid() const
|
||||
{
|
||||
QTextCodec* codec = QTextCodec::codecForName(m_name.c_str());
|
||||
if (codec)
|
||||
if (m_codec)
|
||||
{
|
||||
return true;
|
||||
}
|
||||
@@ -24,23 +26,18 @@ bool TextCodec::isValid() const
|
||||
|
||||
std::wstring TextCodec::decode(const std::string& unicodeString) const
|
||||
{
|
||||
QTextCodec* codec = QTextCodec::codecForName(m_name.c_str());
|
||||
if (codec)
|
||||
if (m_decoder)
|
||||
{
|
||||
QTextDecoder decoder(codec);
|
||||
return decoder.toUnicode(unicodeString.c_str()).toStdWString();
|
||||
return m_decoder->toUnicode(unicodeString.c_str()).toStdWString();
|
||||
}
|
||||
return QString::fromStdString(unicodeString).toStdWString();
|
||||
}
|
||||
|
||||
std::string TextCodec::encode(const std::wstring& string) const
|
||||
{
|
||||
QTextCodec* codec = QTextCodec::codecForName(m_name.c_str());
|
||||
if (codec)
|
||||
if (m_encoder)
|
||||
{
|
||||
QTextEncoder encoder(codec);
|
||||
return encoder.fromUnicode(QString::fromStdWString(string)).toStdString();
|
||||
return m_encoder->fromUnicode(QString::fromStdWString(string)).toStdString();
|
||||
}
|
||||
|
||||
return QString::fromStdWString(string).toStdString();
|
||||
}
|
||||
|
||||
@@ -2,6 +2,11 @@
|
||||
#define TEXT_CODEC_H
|
||||
|
||||
#include <string>
|
||||
#include <memory>
|
||||
|
||||
class QTextCodec;
|
||||
class QTextDecoder;
|
||||
class QTextEncoder;
|
||||
|
||||
class TextCodec
|
||||
{
|
||||
@@ -17,6 +22,9 @@ public:
|
||||
|
||||
private:
|
||||
const std::string m_name;
|
||||
QTextCodec* m_codec;
|
||||
std::shared_ptr<QTextDecoder> m_decoder;
|
||||
std::shared_ptr<QTextEncoder> m_encoder;
|
||||
};
|
||||
|
||||
#endif // TEXT_CODEC_H
|
||||
|
||||
Reference in New Issue
Block a user