1
0
Fork 0
mirror of https://github.com/beefytech/Beef.git synced 2025-07-04 23:36:00 +02:00

Merge pull request #1277 from aharabada/FuzzyAutoComplete

Fuzzy string matiching for autocomplete
This commit is contained in:
Brian Fiete 2021-12-28 17:07:41 +01:00 committed by GitHub
commit 634dd7e509
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
17 changed files with 656 additions and 97 deletions

View file

@ -508,3 +508,11 @@ Path = "X86Target.h"
[[ProjectFolder.Items]]
Type = "Source"
Path = "X86XmmInfo.cpp"
[[ProjectFolder.Items]]
Type = "Folder"
Name = "third_party"
[[ProjectFolder.Items.Items]]
Type = "Source"
Path = "third_party/FtsFuzzyMatch.h"

View file

@ -6,6 +6,9 @@
#include "BfFixits.h"
#include "BfResolvedTypeUtils.h"
#define FTS_FUZZY_MATCH_IMPLEMENTATION
#include "../third_party/FtsFuzzyMatch.h"
#pragma warning(disable:4996)
using namespace llvm;
@ -16,6 +19,7 @@ AutoCompleteBase::AutoCompleteBase()
{
mIsGetDefinition = false;
mIsAutoComplete = true;
mDoFuzzyAutoComplete = false;
mInsertStartIdx = -1;
mInsertEndIdx = -1;
}
@ -25,22 +29,70 @@ AutoCompleteBase::~AutoCompleteBase()
Clear();
}
AutoCompleteEntry* AutoCompleteBase::AddEntry(const AutoCompleteEntry& entry, const StringImpl& filter)
inline void UpdateEntryMatchindices(uint8* matches, AutoCompleteEntry& entry)
{
if ((!DoesFilterMatch(entry.mDisplay, filter.c_str())) || (entry.mNamePrefixCount < 0))
return NULL;
return AddEntry(entry);
if (matches[0] != UINT8_MAX)
{
// Count entries in matches
// Note: entry.mMatchesLength should be the amount of unicode-codepoints in the filter
for (uint8 i = 0;; i++)
{
uint8 matchIndex = matches[i];
if ((matchIndex == 0 && i != 0) || i == UINT8_MAX)
{
entry.mMatchesLength = i;
break;
}
}
entry.mMatches = matches;
}
else
{
entry.mMatches = nullptr;
entry.mMatchesLength = 0;
}
}
AutoCompleteEntry* AutoCompleteBase::AddEntry(const AutoCompleteEntry& entry, const char* filter)
AutoCompleteEntry* AutoCompleteBase::AddEntry(AutoCompleteEntry& entry, const StringImpl& filter)
{
if ((!DoesFilterMatch(entry.mDisplay, filter)) || (entry.mNamePrefixCount < 0))
uint8 matches[256];
if (!DoesFilterMatch(entry.mDisplay, filter.c_str(), entry.mScore, matches, 256) || (entry.mNamePrefixCount < 0))
return NULL;
return AddEntry(entry);
UpdateEntryMatchindices(matches, entry);
auto result = AddEntry(entry);
// Reset matches because the array will be invalid after return
entry.mMatches = nullptr;
entry.mMatchesLength = 0;
return result;
}
AutoCompleteEntry* AutoCompleteBase::AddEntry(AutoCompleteEntry& entry, const char* filter)
{
uint8 matches[256];
if (!DoesFilterMatch(entry.mDisplay, filter, entry.mScore, matches, 256) || (entry.mNamePrefixCount < 0))
return NULL;
UpdateEntryMatchindices(matches, entry);
auto result = AddEntry(entry);
// Reset matches because the array will be invalid after return
entry.mMatches = nullptr;
entry.mMatchesLength = 0;
return result;
}
AutoCompleteEntry* AutoCompleteBase::AddEntry(const AutoCompleteEntry& entry)
{
{
if (mEntriesSet.mAllocSize == 0)
{
mEntriesSet.Reserve(128);
@ -55,13 +107,16 @@ AutoCompleteEntry* AutoCompleteBase::AddEntry(const AutoCompleteEntry& entry)
int size = (int)strlen(display) + 1;
insertedEntry->mDisplay = (char*)mAlloc.AllocBytes(size);
memcpy((char*)insertedEntry->mDisplay, display, size);
insertedEntry->mMatches = (uint8*)mAlloc.AllocBytes(insertedEntry->mMatchesLength);
memcpy((char*)insertedEntry->mMatches, entry.mMatches, insertedEntry->mMatchesLength);
}
return insertedEntry;
}
bool AutoCompleteBase::DoesFilterMatch(const char* entry, const char* filter)
{
bool AutoCompleteBase::DoesFilterMatch(const char* entry, const char* filter, int& score, uint8* matches, int maxMatches)
{
if (mIsGetDefinition)
{
int entryLen = (int)strlen(entry);
@ -73,59 +128,71 @@ bool AutoCompleteBase::DoesFilterMatch(const char* entry, const char* filter)
if (!mIsAutoComplete)
return false;
if (filter[0] == 0)
matches[0] = UINT8_MAX;
if (filter[0] == '\0')
return true;
int filterLen = (int)strlen(filter);
int entryLen = (int)strlen(entry);
bool hasUnderscore = false;
bool checkInitials = filterLen > 1;
for (int i = 0; i < (int)filterLen; i++)
{
char c = filter[i];
if (c == '_')
hasUnderscore = true;
else if (islower((uint8)filter[i]))
checkInitials = false;
}
if (hasUnderscore)
return strnicmp(filter, entry, filterLen) == 0;
char initialStr[256];
char* initialStrP = initialStr;
//String initialStr;
bool prevWasUnderscore = false;
for (int entryIdx = 0; entryIdx < entryLen; entryIdx++)
{
char entryC = entry[entryIdx];
if (entryC == '_')
{
prevWasUnderscore = true;
continue;
}
if ((entryIdx == 0) || (prevWasUnderscore) || (isupper((uint8)entryC) || (isdigit((uint8)entryC))))
{
if (strnicmp(filter, entry + entryIdx, filterLen) == 0)
return true;
if (checkInitials)
*(initialStrP++) = entryC;
}
prevWasUnderscore = false;
if (filterLen == 1)
break; // Don't check inners for single-character case
}
if (!checkInitials)
if (filterLen > entryLen)
return false;
*(initialStrP++) = 0;
return strnicmp(filter, initialStr, filterLen) == 0;
if (mDoFuzzyAutoComplete)
{
return fts::fuzzy_match(filter, entry, score, matches, maxMatches);
}
else
{
bool hasUnderscore = false;
bool checkInitials = filterLen > 1;
for (int i = 0; i < (int)filterLen; i++)
{
char c = filter[i];
if (c == '_')
hasUnderscore = true;
else if (islower((uint8)filter[i]))
checkInitials = false;
}
if (hasUnderscore)
return strnicmp(filter, entry, filterLen) == 0;
char initialStr[256];
char* initialStrP = initialStr;
//String initialStr;
bool prevWasUnderscore = false;
for (int entryIdx = 0; entryIdx < entryLen; entryIdx++)
{
char entryC = entry[entryIdx];
if (entryC == '_')
{
prevWasUnderscore = true;
continue;
}
if ((entryIdx == 0) || (prevWasUnderscore) || (isupper((uint8)entryC) || (isdigit((uint8)entryC))))
{
if (strnicmp(filter, entry + entryIdx, filterLen) == 0)
return true;
if (checkInitials)
*(initialStrP++) = entryC;
}
prevWasUnderscore = false;
if (filterLen == 1)
break; // Don't check inners for single-character case
}
if (!checkInitials)
return false;
*(initialStrP++) = 0;
return strnicmp(filter, initialStr, filterLen) == 0;
}
}
void AutoCompleteBase::Clear()
@ -137,7 +204,7 @@ void AutoCompleteBase::Clear()
//////////////////////////////////////////////////////////////////////////
BfAutoComplete::BfAutoComplete(BfResolveType resolveType)
BfAutoComplete::BfAutoComplete(BfResolveType resolveType, bool doFuzzyAutoComplete)
{
mResolveType = resolveType;
mModule = NULL;
@ -154,6 +221,8 @@ BfAutoComplete::BfAutoComplete(BfResolveType resolveType)
(resolveType == BfResolveType_GoToDefinition);
mIsAutoComplete = (resolveType == BfResolveType_Autocomplete);
mDoFuzzyAutoComplete = doFuzzyAutoComplete;
mGetDefinitionNode = NULL;
mShowAttributeProperties = NULL;
mIdentifierUsed = NULL;
@ -550,7 +619,9 @@ void BfAutoComplete::AddTypeDef(BfTypeDef* typeDef, const StringImpl& filter, bo
return;
}
if (!DoesFilterMatch(name.c_str(), filter.c_str()))
int score;
uint8 matches[256];
if (!DoesFilterMatch(name.c_str(), filter.c_str(), score, matches, sizeof(matches)))
return;
auto type = mModule->ResolveTypeDef(typeDef, BfPopulateType_Declaration);
@ -1128,8 +1199,10 @@ void BfAutoComplete::AddExtensionMethods(BfTypeInstance* targetType, BfTypeInsta
if (methodInstance == NULL)
continue;
int score;
uint8 matches[256];
// Do filter match first- may be cheaper than generic validation
if (!DoesFilterMatch(methodDef->mName.c_str(), filter.c_str()))
if (!DoesFilterMatch(methodDef->mName.c_str(), filter.c_str(), score, matches, sizeof(matches)))
continue;
auto thisType = methodInstance->GetParamType(0);

View file

@ -16,11 +16,16 @@ public:
const char* mDisplay;
const char* mDocumentation;
int8 mNamePrefixCount;
int mScore;
uint8* mMatches;
uint8 mMatchesLength;
public:
AutoCompleteEntry()
{
mNamePrefixCount = 0;
mMatches = nullptr;
mMatchesLength = 0;
}
AutoCompleteEntry(const char* entryType, const char* display)
@ -29,6 +34,9 @@ public:
mDisplay = display;
mDocumentation = NULL;
mNamePrefixCount = 0;
mScore = 0;
mMatches = nullptr;
mMatchesLength = 0;
}
AutoCompleteEntry(const char* entryType, const StringImpl& display)
@ -37,6 +45,9 @@ public:
mDisplay = display.c_str();
mDocumentation = NULL;
mNamePrefixCount = 0;
mScore = 0;
mMatches = nullptr;
mMatchesLength = 0;
}
AutoCompleteEntry(const char* entryType, const StringImpl& display, int namePrefixCount)
@ -45,8 +56,11 @@ public:
mDisplay = display.c_str();
mDocumentation = NULL;
mNamePrefixCount = (int8)namePrefixCount;
mScore = 0;
mMatches = nullptr;
mMatchesLength = 0;
}
bool operator==(const AutoCompleteEntry& other) const
{
return strcmp(mDisplay, other.mDisplay) == 0;
@ -97,12 +111,13 @@ public:
bool mIsGetDefinition;
bool mIsAutoComplete;
bool mDoFuzzyAutoComplete;
int mInsertStartIdx;
int mInsertEndIdx;
bool DoesFilterMatch(const char* entry, const char* filter);
AutoCompleteEntry* AddEntry(const AutoCompleteEntry& entry, const StringImpl& filter);
AutoCompleteEntry* AddEntry(const AutoCompleteEntry& entry, const char* filter);
bool DoesFilterMatch(const char* entry, const char* filter, int& score, uint8* matches, int maxMatches);
AutoCompleteEntry* AddEntry(AutoCompleteEntry& entry, const StringImpl& filter);
AutoCompleteEntry* AddEntry(AutoCompleteEntry& entry, const char* filter);
AutoCompleteEntry* AddEntry(const AutoCompleteEntry& entry);
AutoCompleteBase();
@ -226,7 +241,7 @@ public:
String ConstantToString(BfIRConstHolder* constHolder, BfIRValue id);
public:
BfAutoComplete(BfResolveType resolveType = BfResolveType_Autocomplete);
BfAutoComplete(BfResolveType resolveType = BfResolveType_Autocomplete, bool doFuzzyAutoComplete = false);
~BfAutoComplete();
void SetModule(BfModule* module);

View file

@ -8026,9 +8026,13 @@ void BfCompiler::GenerateAutocompleteInfo()
{
entries.Add(&entry);
}
std::sort(entries.begin(), entries.end(), [](AutoCompleteEntry* lhs, AutoCompleteEntry* rhs)
{
return stricmp(lhs->mDisplay, rhs->mDisplay) < 0;
if (lhs->mScore == rhs->mScore)
return stricmp(lhs->mDisplay, rhs->mDisplay) < 0;
return lhs->mScore > rhs->mScore;
});
String docString;
@ -8043,6 +8047,25 @@ void BfCompiler::GenerateAutocompleteInfo()
autoCompleteResultString += '@';
autoCompleteResultString += String(entry->mDisplay);
if (entry->mMatchesLength > 0)
{
autoCompleteResultString += "\x02";
for (int i = 0; i < entry->mMatchesLength; i++)
{
int match = entry->mMatches[i];
// Need max 3 chars (largest Hex (FF) + '\0')
char buffer[3];
_itoa_s(match, buffer, 16);
autoCompleteResultString += String(buffer);
autoCompleteResultString += ",";
}
autoCompleteResultString += "X";
}
if (entry->mDocumentation != NULL)
{
autoCompleteResultString += '\x03';

View file

@ -2071,7 +2071,7 @@ void BfModule::UpdateCEEmit(CeEmitContext* ceEmitContext, BfTypeInstance* typeIn
{
for (int ifaceTypeId : ceEmitContext->mInterfaces)
typeInstance->mCeTypeInfo->mPendingInterfaces.Add(ifaceTypeId);
if (ceEmitContext->mEmitData.IsEmpty())
return;

View file

@ -3898,13 +3898,13 @@ BF_EXPORT const char* BF_CALLTYPE BfParser_GetDebugExpressionAt(BfParser* bfPars
return outString.c_str();
}
BF_EXPORT BfResolvePassData* BF_CALLTYPE BfParser_CreateResolvePassData(BfParser* bfParser, BfResolveType resolveType)
BF_EXPORT BfResolvePassData* BF_CALLTYPE BfParser_CreateResolvePassData(BfParser* bfParser, BfResolveType resolveType, bool doFuzzyAutoComplete)
{
auto bfResolvePassData = new BfResolvePassData();
bfResolvePassData->mResolveType = resolveType;
bfResolvePassData->mParser = bfParser;
if ((bfParser != NULL) && ((bfParser->mParserFlags & ParserFlag_Autocomplete) != 0))
bfResolvePassData->mAutoComplete = new BfAutoComplete(resolveType);
bfResolvePassData->mAutoComplete = new BfAutoComplete(resolveType, doFuzzyAutoComplete);
return bfResolvePassData;
}

View file

@ -400,6 +400,7 @@
<ClInclude Include="Compiler\BfUtil.h" />
<ClInclude Include="Compiler\BfVarDeclChecker.h" />
<ClInclude Include="Compiler\CeMachine.h" />
<ClInclude Include="third_party\FtsFuzzyMatch.h" />
<ClInclude Include="Compiler\MemReporter.h" />
<ClInclude Include="DbgMiniDump.h" />
<ClInclude Include="Debugger.h" />

View file

@ -24,6 +24,9 @@
<Filter Include="Beef">
<UniqueIdentifier>{83b97406-2f83-49ad-bbbc-3ff70ecda6bb}</UniqueIdentifier>
</Filter>
<Filter Include="third_party">
<UniqueIdentifier>{d36777f2-b326-4a8c-84a3-5c2f39153f75}</UniqueIdentifier>
</Filter>
</ItemGroup>
<ItemGroup>
<ClCompile Include="Compiler\BfAst.cpp">
@ -399,5 +402,8 @@
<ClInclude Include="Compiler\CeMachine.h">
<Filter>Compiler</Filter>
</ClInclude>
<ClInclude Include="third_party\FtsFuzzyMatch.h">
<Filter>third_party</Filter>
</ClInclude>
</ItemGroup>
</Project>

View file

@ -183,7 +183,7 @@ namespace Tests
mStr.AppendF($"{name} {val}\n");
}
}
interface ISerializable
{
void Serialize(SerializationContext ctx);

256
IDEHelper/third_party/FtsFuzzyMatch.h vendored Normal file
View file

@ -0,0 +1,256 @@
// LICENSE
//
// This software is dual-licensed to the public domain and under the following
// license: you are granted a perpetual, irrevocable license to copy, modify,
// publish, and distribute this file as you see fit.
//
// VERSION
// 0.2.0 (2017-02-18) Scored matches perform exhaustive search for best score
// 0.1.0 (2016-03-28) Initial release
//
// AUTHOR
// Forrest Smith
//
// NOTES
// Compiling
// You MUST add '#define FTS_FUZZY_MATCH_IMPLEMENTATION' before including this header in ONE source file to create implementation.
//
// fuzzy_match_simple(...)
// Returns true if each character in pattern is found sequentially within str
//
// fuzzy_match(...)
// Returns true if pattern is found AND calculates a score.
// Performs exhaustive search via recursion to find all possible matches and match with highest score.
// Scores values have no intrinsic meaning. Possible score range is not normalized and varies with pattern.
// Recursion is limited internally (default=10) to prevent degenerate cases (pattern="aaaaaa" str="aaaaaaaaaaaaaaaaaaaaaaaaaaaaaa")
// Uses uint8_t for match indices. Therefore patterns are limited to 256 characters.
// Score system should be tuned for YOUR use case. Words, sentences, file names, or method names all prefer different tuning.
#ifndef FTS_FUZZY_MATCH_H
#define FTS_FUZZY_MATCH_H
#include <cstdint> // uint8_t
#include <ctype.h> // ::tolower, ::toupper
#include <cstring> // memcpy
#include <cstdio>
#include "BeefySysLib/util/UTF8.h"
#include "BeefySysLib/third_party/utf8proc/utf8proc.h"
// Public interface
namespace fts {
static bool fuzzy_match_simple(char const* pattern, char const* str);
static bool fuzzy_match(char const* pattern, char const* str, int& outScore);
static bool fuzzy_match(char const* pattern, char const* str, int& outScore, uint8_t* matches, int maxMatches);
}
BF_EXPORT bool BF_CALLTYPE fts_fuzzy_match(char const* pattern, char const* str, int& outScore, uint8_t* matches, int maxMatches);
#ifdef FTS_FUZZY_MATCH_IMPLEMENTATION
namespace fts {
// Forward declarations for "private" implementation
namespace fuzzy_internal {
static bool fuzzy_match_recursive(const char* pattern, const char* str, int& outScore, const char* strBegin,
uint8_t const* srcMatches, uint8_t* newMatches, int maxMatches, int nextMatch,
int& recursionCount, int recursionLimit);
}
// Public interface
static bool fuzzy_match_simple(char const* pattern, char const* str) {
while (*pattern != '\0' && *str != '\0') {
if (tolower(*pattern) == tolower(*str))
++pattern;
++str;
}
return *pattern == '\0' ? true : false;
}
static bool fuzzy_match(char const* pattern, char const* str, int& outScore) {
uint8_t matches[256];
return fuzzy_match(pattern, str, outScore, matches, sizeof(matches));
}
static bool fuzzy_match(char const* pattern, char const* str, int& outScore, uint8_t* matches, int maxMatches) {
int recursionCount = 0;
int recursionLimit = 10;
return fuzzy_internal::fuzzy_match_recursive(pattern, str, outScore, str, nullptr, matches, maxMatches, 0, recursionCount, recursionLimit);
}
bool IsLower(uint32 c)
{
return utf8proc_category(c) == UTF8PROC_CATEGORY_LL;
}
bool IsUpper(uint32 c)
{
return utf8proc_category(c) == UTF8PROC_CATEGORY_LU;
}
// Private implementation
static bool fuzzy_internal::fuzzy_match_recursive(const char* pattern, const char* str, int& outScore,
const char* strBegin, uint8_t const* srcMatches, uint8_t* matches, int maxMatches,
int nextMatch, int& recursionCount, int recursionLimit)
{
// Count recursions
++recursionCount;
if (recursionCount >= recursionLimit)
return false;
// Detect end of strings
if (*pattern == '\0' || *str == '\0')
return false;
// Recursion params
bool recursiveMatch = false;
uint8_t bestRecursiveMatches[256];
int bestRecursiveScore = 0;
// Loop through pattern and str looking for a match
bool first_match = true;
while (*pattern != '\0' && *str != '\0') {
int patternOffset = 0;
uint32 patternChar = Beefy::u8_nextchar((char*)pattern, &patternOffset);
int strOffset = 0;
uint32 strChar = Beefy::u8_nextchar((char*)str, &strOffset);
// TODO: tolower only works for A-Z
// Found match
if (utf8proc_tolower(patternChar) == utf8proc_tolower(strChar)) {
// Supplied matches buffer was too short
if (nextMatch >= maxMatches)
return false;
// "Copy-on-Write" srcMatches into matches
if (first_match && srcMatches) {
memcpy(matches, srcMatches, nextMatch);
first_match = false;
}
// Recursive call that "skips" this match
uint8_t recursiveMatches[256];
int recursiveScore;
if (fuzzy_match_recursive(pattern, str + strOffset, recursiveScore, strBegin, matches, recursiveMatches, sizeof(recursiveMatches), nextMatch, recursionCount, recursionLimit)) {
// Pick best recursive score
if (!recursiveMatch || recursiveScore > bestRecursiveScore) {
memcpy(bestRecursiveMatches, recursiveMatches, 256);
bestRecursiveScore = recursiveScore;
}
recursiveMatch = true;
}
// Advance
matches[nextMatch++] = (uint8_t)(str - strBegin);
// Clear the next char so that we know which match is the last one
matches[nextMatch + 1] = 0;
pattern += patternOffset;
}
str += strOffset;
}
// Determine if full pattern was matched
bool matched = *pattern == '\0' ? true : false;
// Calculate score
if (matched) {
const int sequential_bonus = 15; // bonus for adjacent matches
const int separator_bonus = 30; // bonus if match occurs after a separator
const int camel_bonus = 30; // bonus if match is uppercase and prev is lower
const int first_letter_bonus = 15; // bonus if the first letter is matched
const int leading_letter_penalty = -5; // penalty applied for every letter in str before the first match
const int max_leading_letter_penalty = -15; // maximum penalty for leading letters
const int unmatched_letter_penalty = -1; // penalty for every letter that doesn't matter
// Iterate str to end
while (*str != '\0')
++str;
// Initialize score
outScore = 100;
// Apply leading letter penalty
int penalty = leading_letter_penalty * matches[0];
if (penalty < max_leading_letter_penalty)
penalty = max_leading_letter_penalty;
outScore += penalty;
// Apply unmatched penalty
int unmatched = (int)(str - strBegin) - nextMatch;
outScore += unmatched_letter_penalty * unmatched;
// Apply ordering bonuses
for (int i = 0; i < nextMatch; ++i) {
uint8_t currIdx = matches[i];
int currOffset = currIdx;
uint32 curr = Beefy::u8_nextchar((char*)strBegin, &currOffset);
if (i > 0) {
uint8_t prevIdx = matches[i - 1];
int offsetPrevidx = prevIdx;
Beefy::u8_inc((char*)strBegin, &offsetPrevidx);
// Sequential
if (currIdx == offsetPrevidx)
outScore += sequential_bonus;
}
// Check for bonuses based on neighbor character value
if (currIdx > 0) {
int neighborOffset = currIdx;
Beefy::u8_dec((char*)strBegin, &neighborOffset);
uint32 neighbor = Beefy::u8_nextchar((char*)strBegin, &neighborOffset);
// Camel case
if (IsLower(neighbor) && IsUpper(curr))
outScore += camel_bonus;
// Separator
bool neighborSeparator = neighbor == '_' || neighbor == ' ';
if (neighborSeparator)
outScore += separator_bonus;
}
else {
// First letter
outScore += first_letter_bonus;
}
}
}
// Return best result
if (recursiveMatch && (!matched || bestRecursiveScore > outScore)) {
// Recursive score is better than "this"
memcpy(matches, bestRecursiveMatches, maxMatches);
outScore = bestRecursiveScore;
return true;
}
else if (matched) {
// "this" score is better than recursive
return true;
}
else {
// no match
return false;
}
}
} // namespace fts
BF_EXPORT bool BF_CALLTYPE fts_fuzzy_match(char const* pattern, char const* str, int& outScore, uint8_t* matches, int maxMatches)
{
return fts::fuzzy_match(pattern, str, outScore, matches, maxMatches);
}
#endif // FTS_FUZZY_MATCH_IMPLEMENTATION
#endif // FTS_FUZZY_MATCH_H