Merge pull request #18748 from dcamper/hpcc-31921-regex-caching-master

HPCC-31921 Add caching of regex compiled search patterns Reviewed-by: Gavin Halliday <[email protected]> Merged-by: Gavin Halliday <[email protected]>
hpcc-systems · Jun 20, 2024 · 683bc67 · 683bc67
2 parents b4d06e3 + 7997a0a
commit 683bc67
Show file tree

Hide file tree

Showing 11 changed files with 677 additions and 101 deletions.
diff --git a/ecl/hql/hqlfold.cpp b/ecl/hql/hqlfold.cpp
@@ -2796,7 +2796,7 @@ IHqlExpression * foldConstantOperator(IHqlExpression * expr, unsigned foldOption
                     StringBuffer pattern, search;
                     v0->getUTF8Value(pattern);
                     v1->getUTF8Value(search);
-                    ICompiledStrRegExpr * compiled = rtlCreateCompiledU8StrRegExpr(pattern, !expr->hasAttribute(noCaseAtom));
+                    ICompiledStrRegExpr * compiled = rtlCreateCompiledU8StrRegExpr(pattern.lengthUtf8(), pattern, !expr->hasAttribute(noCaseAtom));
                     compiled->getMatchSet(isAllResult, resultBytes, matchResults.refdata(), search.lengthUtf8(), search.str());
                     rtlDestroyCompiledU8StrRegExpr(compiled);
                 }

diff --git a/rtl/eclrtl/eclregex.cpp b/rtl/eclrtl/eclregex.cpp
diff --git a/rtl/eclrtl/eclrtl.cpp b/rtl/eclrtl/eclrtl.cpp
@@ -4960,6 +4960,19 @@ unsigned rtlUtf8Size(unsigned len, const void * _data)
     return offset;
 }
 
+unsigned rtlUtf8Length(const void * _data)
+{
+    const byte * data = (const byte *)_data;
+    size32_t length = 0;
+    unsigned offset = 0;
+    while (data[offset])
+    {
+        offset += readUtf8Size(data+offset);
+        length++;
+    }
+    return length;
+}
+
 unsigned rtlUtf8Length(unsigned size, const void * _data)
 {
     const byte * data = (const byte *)_data;

diff --git a/rtl/eclrtl/eclrtl.hpp b/rtl/eclrtl/eclrtl.hpp
@@ -667,6 +667,7 @@ ECLRTL_API void rtlStrToVUnicode(unsigned outlen, UChar * out, unsigned inlen, c
 
 ECLRTL_API unsigned rtlUtf8Size(const void * data);
 ECLRTL_API unsigned rtlUtf8Size(unsigned len, const void * data);
+ECLRTL_API unsigned rtlUtf8Length(const void * data);
 ECLRTL_API unsigned rtlUtf8Length(unsigned size, const void * data);
 ECLRTL_API unsigned rtlUtf8Char(const void * data);
 ECLRTL_API void rtlUtf8ToData(size32_t outlen, void * out, size32_t inlen, const char *in);

diff --git a/system/jlib/jhash.hpp b/system/jlib/jhash.hpp
@@ -21,6 +21,7 @@
 #define JHASH_HPP
 
 #include <functional>
+#include <list>
 #include <unordered_map>
 #include <utility>
 
@@ -687,4 +688,103 @@ class CTimeLimitedCache
     }
 };
 
+/**
+ * CLRUCache
+ *
+ * Least-Recently-Used cache class, specialized for key and
+ * value pointer types (the value is a pointer or a data type
+ * where a nullptr could represent a missing value).
+ *
+ * The get() method returns a found object by value.  This
+ * is intentional and very useful for maintaining refcounts.
+ *
+ * There is a minimum size for the cache, defined by
+ * LRU_MIN_CACHE_SIZE.  Attempts to create a smaller cache
+ * will be silently changed to the minimum size.  If no
+ * initial size is provided to the constructor, the cache
+ * size will be set to LRU_MIN_CACHE_SIZE.
+ *
+ * Methods here are not thread-safe.  Callers should block
+ * concurrent access for non-const methods (which are most
+ * of them).
+ */
+
+#define LRU_MIN_CACHE_SIZE 10
+
+template <class KEYTYPE, class PTRTYPE>
+class CLRUCache
+{
+    private:
+        std::list<KEYTYPE> recentList;
+        std::unordered_map<KEYTYPE, std::pair<PTRTYPE, typename std::list<KEYTYPE>::iterator>> lookupMap;
+        size32_t maxCacheSize;
+
+        void _downsize()
+        {
+            while (lookupMap.size() > maxCacheSize)
+            {
+                lookupMap.erase(recentList.back());
+                recentList.pop_back();
+            }
+        }
+
+    public:
+        CLRUCache() : maxCacheSize(LRU_MIN_CACHE_SIZE) {}
+        CLRUCache(size32_t _maxCacheSize) : maxCacheSize(_maxCacheSize < LRU_MIN_CACHE_SIZE ? LRU_MIN_CACHE_SIZE : _maxCacheSize) {}
+        CLRUCache(const CLRUCache& other) = delete;
+        ~CLRUCache() = default;
+
+        size32_t getCacheSize() const
+        {
+            return lookupMap.size();
+        }
+
+        size32_t setMaxCacheSize(size32_t _maxCacheSize)
+        {
+            maxCacheSize = _maxCacheSize < LRU_MIN_CACHE_SIZE ? LRU_MIN_CACHE_SIZE : _maxCacheSize;
+            _downsize();
+            return maxCacheSize;
+        }
+
+        PTRTYPE get(const KEYTYPE& key)
+        {
+            auto foundIter = lookupMap.find(key);
+            if (foundIter == lookupMap.end())
+                return nullptr;
+
+            recentList.splice(recentList.begin(), recentList, foundIter->second.second);
+            foundIter->second.second = recentList.begin();
+            return foundIter->second.first;
+        }
+
+        void set(const KEYTYPE& key, const PTRTYPE& value)
+        {
+            auto foundIter = lookupMap.find(key);
+
+            if (foundIter == lookupMap.end())
+            {
+                recentList.push_front(key);
+                lookupMap[key] = {value, recentList.begin()};
+                _downsize();
+            }
+            else
+            {
+                recentList.splice(recentList.begin(), recentList, foundIter->second.second);
+                foundIter->second.first = value;
+                foundIter->second.second = recentList.begin();
+            }
+        }
+
+        bool remove(const KEYTYPE& key)
+        {
+            auto foundIter = lookupMap.find(key);
+            if (foundIter == lookupMap.end())
+                return false;
+
+            recentList.erase(foundIter->second.second);
+            lookupMap.erase(foundIter);
+            return true;
+        }
+};
+
 #endif
diff --git a/testing/regress/ecl/key/regex_cache_string.xml b/testing/regress/ecl/key/regex_cache_string.xml
@@ -0,0 +1,3 @@
+<Dataset name='result'>
+ <Row><result>PASSED</result></Row>
+</Dataset>
diff --git a/testing/regress/ecl/key/regex_cache_unicode.xml b/testing/regress/ecl/key/regex_cache_unicode.xml
@@ -0,0 +1,3 @@
+<Dataset name='result'>
+ <Row><result>PASSED</result></Row>
+</Dataset>
diff --git a/testing/regress/ecl/key/regex_cache_utf8.xml b/testing/regress/ecl/key/regex_cache_utf8.xml
@@ -0,0 +1,3 @@
+<Dataset name='result'>
+ <Row><result>PASSED</result></Row>
+</Dataset>
diff --git a/testing/regress/ecl/regex_cache_string.ecl b/testing/regress/ecl/regex_cache_string.ecl
@@ -0,0 +1,70 @@
+/*##############################################################################
+
+    HPCC SYSTEMS software Copyright (C) 2024 HPCC Systems®.
+    Licensed under the Apache License, Version 2.0 (the "License");
+    you may not use this file except in compliance with the License.
+    You may obtain a copy of the License at
+       http://www.apache.org/licenses/LICENSE-2.0
+    Unless required by applicable law or agreed to in writing, software
+    distributed under the License is distributed on an "AS IS" BASIS,
+    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+    See the License for the specific language governing permissions and
+    limitations under the License.
+############################################################################## */
+
+#OPTION('globalFold', FALSE);
+
+IMPORT Std;
+
+regexDS := DATASET
+    (
+        100000,
+        TRANSFORM
+            (
+                {STRING a},
+                SELF.a := (STRING)RANDOM()
+            ),
+        DISTRIBUTED
+    );
+
+res := PROJECT
+    (
+        NOFOLD(regexDS),
+        TRANSFORM
+            (
+                {
+                    RECORDOF(LEFT),
+                    STRING via_regex,
+                    STRING via_find,
+                    BOOLEAN is_matching
+                },
+                SELF.via_regex := REGEXREPLACE(LEFT.a[1], LEFT.a, 'x'),
+                SELF.via_find := Std.Str.SubstituteIncluded(LEFT.a, LEFT.a[1], 'x'),
+                SELF.is_matching := SELF.via_regex = SELF.via_find,
+                SELF := LEFT
+            ),
+        PARALLEL(10)
+    );
+
+numTests := COUNT(regexDS);
+testsPassed := res(is_matching);
+numTestsPassed := COUNT(testsPassed);
+testsFailed := res(~is_matching);
+numTestsFailed := COUNT(testsFailed);
+
+MIN_PASS_PERCENTAGE := 0.95;
+
+passedPercentage := numTestsPassed / numTests;
+isSuccess := passedPercentage >= MIN_PASS_PERCENTAGE;
+resultStr := IF(isSuccess, 'PASSED', 'FAILED');
+fullResultStr := resultStr + ': ' + (STRING)(ROUND(passedPercentage * 100, 2));
+
+// Output for unit test parsing
+OUTPUT(resultStr, NAMED('result'));
+
+// Uncomment the following to see details
+// OUTPUT(numTests, NAMED('num_tests'));
+// OUTPUT(numTestsPassed, NAMED('num_passed'));
+// OUTPUT(numTestsFailed, NAMED('num_failed'));
+// OUTPUT(fullResultStr, NAMED('result_desc'));
+// OUTPUT(testsFailed, NAMED('failed_tests'), ALL);
diff --git a/testing/regress/ecl/regex_cache_unicode.ecl b/testing/regress/ecl/regex_cache_unicode.ecl
@@ -0,0 +1,70 @@
+/*##############################################################################
+
+    HPCC SYSTEMS software Copyright (C) 2024 HPCC Systems®.
+    Licensed under the Apache License, Version 2.0 (the "License");
+    you may not use this file except in compliance with the License.
+    You may obtain a copy of the License at
+       http://www.apache.org/licenses/LICENSE-2.0
+    Unless required by applicable law or agreed to in writing, software
+    distributed under the License is distributed on an "AS IS" BASIS,
+    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+    See the License for the specific language governing permissions and
+    limitations under the License.
+############################################################################## */
+
+#OPTION('globalFold', FALSE);
+
+IMPORT Std;
+
+regexDS := DATASET
+    (
+        100000,
+        TRANSFORM
+            (
+                {UNICODE a},
+                SELF.a := (UNICODE)RANDOM()
+            ),
+        DISTRIBUTED
+    );
+
+res := PROJECT
+    (
+        NOFOLD(regexDS),
+        TRANSFORM
+            (
+                {
+                    RECORDOF(LEFT),
+                    UNICODE via_regex,
+                    UNICODE via_find,
+                    BOOLEAN is_matching
+                },
+                SELF.via_regex := REGEXREPLACE(LEFT.a[1], LEFT.a, u'x'),
+                SELF.via_find := (UNICODE)Std.Uni.SubstituteIncluded(LEFT.a, LEFT.a[1], u'x'),
+                SELF.is_matching := SELF.via_regex = SELF.via_find,
+                SELF := LEFT
+            ),
+        PARALLEL(10)
+    );
+
+numTests := COUNT(regexDS);
+testsPassed := res(is_matching);
+numTestsPassed := COUNT(testsPassed);
+testsFailed := res(~is_matching);
+numTestsFailed := COUNT(testsFailed);
+
+MIN_PASS_PERCENTAGE := 0.95;
+
+passedPercentage := numTestsPassed / numTests;
+isSuccess := passedPercentage >= MIN_PASS_PERCENTAGE;
+resultStr := IF(isSuccess, 'PASSED', 'FAILED');
+fullResultStr := resultStr + ': ' + (STRING)(ROUND(passedPercentage * 100, 2));
+
+// Output for unit test parsing
+OUTPUT(resultStr, NAMED('result'));
+
+// Uncomment the following to see details
+// OUTPUT(numTests, NAMED('num_tests'));
+// OUTPUT(numTestsPassed, NAMED('num_passed'));
+// OUTPUT(numTestsFailed, NAMED('num_failed'));
+// OUTPUT(fullResultStr, NAMED('result_desc'));
+// OUTPUT(testsFailed, NAMED('failed_tests'), ALL);
diff --git a/testing/regress/ecl/regex_cache_utf8.ecl b/testing/regress/ecl/regex_cache_utf8.ecl
@@ -0,0 +1,70 @@
+/*##############################################################################
+
+    HPCC SYSTEMS software Copyright (C) 2024 HPCC Systems®.
+    Licensed under the Apache License, Version 2.0 (the "License");
+    you may not use this file except in compliance with the License.
+    You may obtain a copy of the License at
+       http://www.apache.org/licenses/LICENSE-2.0
+    Unless required by applicable law or agreed to in writing, software
+    distributed under the License is distributed on an "AS IS" BASIS,
+    WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+    See the License for the specific language governing permissions and
+    limitations under the License.
+############################################################################## */
+
+#OPTION('globalFold', FALSE);
+
+IMPORT Std;
+
+regexDS := DATASET
+    (
+        100000,
+        TRANSFORM
+            (
+                {UTF8 a},
+                SELF.a := (UTF8)RANDOM()
+            ),
+        DISTRIBUTED
+    );
+
+res := PROJECT
+    (
+        NOFOLD(regexDS),
+        TRANSFORM
+            (
+                {
+                    RECORDOF(LEFT),
+                    UTF8 via_regex,
+                    UTF8 via_find,
+                    BOOLEAN is_matching
+                },
+                SELF.via_regex := REGEXREPLACE(LEFT.a[1], LEFT.a, u8'x'),
+                SELF.via_find := (UTF8)Std.Uni.SubstituteIncluded(LEFT.a, LEFT.a[1], u8'x'),
+                SELF.is_matching := SELF.via_regex = SELF.via_find,
+                SELF := LEFT
+            ),
+        PARALLEL(10)
+    );
+
+numTests := COUNT(regexDS);
+testsPassed := res(is_matching);
+numTestsPassed := COUNT(testsPassed);
+testsFailed := res(~is_matching);
+numTestsFailed := COUNT(testsFailed);
+
+MIN_PASS_PERCENTAGE := 0.95;
+
+passedPercentage := numTestsPassed / numTests;
+isSuccess := passedPercentage >= MIN_PASS_PERCENTAGE;
+resultStr := IF(isSuccess, 'PASSED', 'FAILED');
+fullResultStr := resultStr + ': ' + (STRING)(ROUND(passedPercentage * 100, 2));
+
+// Output for unit test parsing
+OUTPUT(resultStr, NAMED('result'));
+
+// Uncomment the following to see details
+// OUTPUT(numTests, NAMED('num_tests'));
+// OUTPUT(numTestsPassed, NAMED('num_passed'));
+// OUTPUT(numTestsFailed, NAMED('num_failed'));
+// OUTPUT(fullResultStr, NAMED('result_desc'));
+// OUTPUT(testsFailed, NAMED('failed_tests'), ALL);