ClickHouse/src/Functions/MatchImpl.h
2023-09-14 16:40:42 +00:00

685 lines
25 KiB
C++

#pragma once
#include <type_traits>
#include <base/types.h>
#include <Common/Volnitsky.h>
#include <Columns/ColumnString.h>
#include <Columns/ColumnsNumber.h>
#include <Core/ColumnNumbers.h>
#include <Functions/Regexps.h>
#include "config.h"
#include <re2/re2.h>
namespace DB
{
namespace ErrorCodes
{
extern const int ILLEGAL_COLUMN;
}
namespace impl
{
/// Is the [I]LIKE expression equivalent to a substring search?
inline bool likePatternIsSubstring(std::string_view pattern, String & res)
{
/// TODO: ignore multiple leading or trailing %
if (pattern.size() < 2 || !pattern.starts_with('%') || !pattern.ends_with('%'))
return false;
res.clear();
res.reserve(pattern.size() - 2);
const char * pos = pattern.data() + 1;
const char * const end = pattern.data() + pattern.size() - 1;
while (pos < end)
{
switch (*pos)
{
case '%':
case '_':
return false;
case '\\':
++pos;
if (pos == end)
/// pattern ends with \% --> trailing % is to be taken literally and pattern doesn't qualify for substring search
return false;
else
{
switch (*pos)
{
/// Known LIKE escape sequences:
case '%':
case '_':
case '\\':
res += *pos;
break;
/// For all other escape sequences, the backslash loses its special meaning
default:
res += '\\';
res += *pos;
break;
}
}
break;
default:
res += *pos;
break;
}
++pos;
}
return true;
}
}
// For more readable instantiations of MatchImpl<>
struct MatchTraits
{
enum class Syntax
{
Like,
Re2
};
enum class Case
{
Sensitive,
Insensitive
};
enum class Result
{
DontNegate,
Negate
};
};
/**
* NOTE: We want to run regexp search for whole columns by one call (as implemented in function 'position')
* but for that, regexp engine must support \0 bytes and their interpretation as string boundaries.
*/
template <typename Name, MatchTraits::Syntax syntax_, MatchTraits::Case case_, MatchTraits::Result result_>
struct MatchImpl
{
static constexpr bool use_default_implementation_for_constants = true;
static constexpr bool supports_start_pos = false;
static constexpr auto name = Name::name;
static ColumnNumbers getArgumentsThatAreAlwaysConstant() { return {2};}
using ResultType = UInt8;
static constexpr bool is_like = (syntax_ == MatchTraits::Syntax::Like);
static constexpr bool case_insensitive = (case_ == MatchTraits::Case::Insensitive);
static constexpr bool negate = (result_ == MatchTraits::Result::Negate);
using Searcher = std::conditional_t<case_insensitive, VolnitskyCaseInsensitiveUTF8, VolnitskyUTF8>;
static void vectorConstant(
const ColumnString::Chars & haystack_data,
const ColumnString::Offsets & haystack_offsets,
const String & needle,
[[maybe_unused]] const ColumnPtr & start_pos_,
PaddedPODArray<UInt8> & res,
[[maybe_unused]] ColumnUInt8 * res_null)
{
/// `res_null` serves as an output parameter for implementing an XYZOrNull variant.
assert(!res_null);
const size_t haystack_size = haystack_offsets.size();
assert(haystack_size == res.size());
assert(start_pos_ == nullptr);
if (haystack_offsets.empty())
return;
/// Shortcut for the silly but practical case that the pattern matches everything/nothing independently of the haystack:
/// - col [not] [i]like '%' / '%%'
/// - match(col, '.*')
if ((is_like && (needle == "%" or needle == "%%")) || (!is_like && (needle == ".*" || needle == ".*?")))
{
for (auto & x : res)
x = !negate;
return;
}
/// Special case that the [I]LIKE expression reduces to finding a substring in a string
String strstr_pattern;
if (is_like && impl::likePatternIsSubstring(needle, strstr_pattern))
{
const UInt8 * const begin = haystack_data.data();
const UInt8 * const end = haystack_data.data() + haystack_data.size();
const UInt8 * pos = begin;
/// The current index in the array of strings.
size_t i = 0;
/// TODO You need to make that `searcher` is common to all the calls of the function.
Searcher searcher(strstr_pattern.data(), strstr_pattern.size(), end - pos);
/// We will search for the next occurrence in all rows at once.
while (pos < end && end != (pos = searcher.search(pos, end - pos)))
{
/// Let's determine which index it refers to.
while (begin + haystack_offsets[i] <= pos)
{
res[i] = negate;
++i;
}
/// We check that the entry does not pass through the boundaries of strings.
if (pos + strstr_pattern.size() < begin + haystack_offsets[i])
res[i] = !negate;
else
res[i] = negate;
pos = begin + haystack_offsets[i];
++i;
}
/// Tail, in which there can be no substring.
if (i < res.size())
memset(&res[i], negate, (res.size() - i) * sizeof(res[0]));
return;
}
const auto & regexp = OptimizedRegularExpression(Regexps::createRegexp<is_like, /*no_capture*/ true, case_insensitive>(needle));
String required_substring;
bool is_trivial;
bool required_substring_is_prefix; /// for `anchored` execution of the regexp.
regexp.getAnalyzeResult(required_substring, is_trivial, required_substring_is_prefix);
if (required_substring.empty())
{
if (!regexp.getRE2()) /// An empty regexp. Always matches.
memset(res.data(), !negate, haystack_size * sizeof(res[0]));
else
{
size_t prev_offset = 0;
for (size_t i = 0; i < haystack_size; ++i)
{
const bool match = regexp.getRE2()->Match(
{reinterpret_cast<const char *>(&haystack_data[prev_offset]), haystack_offsets[i] - prev_offset - 1},
0,
haystack_offsets[i] - prev_offset - 1,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match;
prev_offset = haystack_offsets[i];
}
}
}
else
{
/// NOTE This almost matches with the case of impl::likePatternIsSubstring.
const UInt8 * const begin = haystack_data.data();
const UInt8 * const end = haystack_data.begin() + haystack_data.size();
const UInt8 * pos = begin;
/// The current index in the array of strings.
size_t i = 0;
Searcher searcher(required_substring.data(), required_substring.size(), end - pos);
/// We will search for the next occurrence in all rows at once.
while (pos < end && end != (pos = searcher.search(pos, end - pos)))
{
/// Determine which index it refers to.
while (begin + haystack_offsets[i] <= pos)
{
res[i] = negate;
++i;
}
/// We check that the entry does not pass through the boundaries of strings.
if (pos + required_substring.size() < begin + haystack_offsets[i])
{
/// And if it does not, if necessary, we check the regexp.
if (is_trivial)
res[i] = !negate;
else
{
const char * str_data = reinterpret_cast<const char *>(&haystack_data[haystack_offsets[i - 1]]);
size_t str_size = haystack_offsets[i] - haystack_offsets[i - 1] - 1;
/** Even in the case of `required_substring_is_prefix` use UNANCHORED check for regexp,
* so that it can match when `required_substring` occurs into the string several times,
* and at the first occurrence, the regexp is not a match.
*/
const size_t start_pos = (required_substring_is_prefix) ? (reinterpret_cast<const char *>(pos) - str_data) : 0;
const size_t end_pos = str_size;
const bool match = regexp.getRE2()->Match(
{str_data, str_size},
start_pos,
end_pos,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match;
}
}
else
res[i] = negate;
pos = begin + haystack_offsets[i];
++i;
}
/// Tail, in which there can be no substring.
if (i < res.size())
memset(&res[i], negate, (res.size() - i) * sizeof(res[0]));
}
}
/// Very carefully crafted copy-paste.
static void vectorFixedConstant(
const ColumnString::Chars & haystack,
size_t N,
const String & needle,
PaddedPODArray<UInt8> & res,
[[maybe_unused]] ColumnUInt8 * res_null)
{
/// `res_null` serves as an output parameter for implementing an XYZOrNull variant.
assert(!res_null);
const size_t haystack_size = haystack.size() / N;
assert(haystack_size == res.size());
if (haystack.empty())
return;
/// Shortcut for the silly but practical case that the pattern matches everything/nothing independently of the haystack:
/// - col [not] [i]like '%' / '%%'
/// - match(col, '.*')
if ((is_like && (needle == "%" or needle == "%%")) || (!is_like && (needle == ".*" || needle == ".*?")))
{
for (auto & x : res)
x = !negate;
return;
}
/// Special case that the [I]LIKE expression reduces to finding a substring in a string
String strstr_pattern;
if (is_like && impl::likePatternIsSubstring(needle, strstr_pattern))
{
const UInt8 * const begin = haystack.data();
const UInt8 * const end = haystack.data() + haystack.size();
const UInt8 * pos = begin;
size_t i = 0;
const UInt8 * next_pos = begin;
/// If needle is larger than string size - it cannot be found.
if (strstr_pattern.size() <= N)
{
Searcher searcher(strstr_pattern.data(), strstr_pattern.size(), end - pos);
/// We will search for the next occurrence in all rows at once.
while (pos < end && end != (pos = searcher.search(pos, end - pos)))
{
/// Let's determine which index it refers to.
while (next_pos + N <= pos)
{
res[i] = negate;
next_pos += N;
++i;
}
next_pos += N;
/// We check that the entry does not pass through the boundaries of strings.
if (pos + strstr_pattern.size() <= next_pos)
res[i] = !negate;
else
res[i] = negate;
pos = next_pos;
++i;
}
}
/// Tail, in which there can be no substring.
if (i < res.size())
memset(&res[i], negate, (res.size() - i) * sizeof(res[0]));
return;
}
const auto & regexp = OptimizedRegularExpression(Regexps::createRegexp<is_like, /*no_capture*/ true, case_insensitive>(needle));
String required_substring;
bool is_trivial;
bool required_substring_is_prefix; /// for `anchored` execution of the regexp.
regexp.getAnalyzeResult(required_substring, is_trivial, required_substring_is_prefix);
if (required_substring.empty())
{
if (!regexp.getRE2()) /// An empty regexp. Always matches.
memset(res.data(), !negate, haystack_size * sizeof(res[0]));
else
{
size_t offset = 0;
for (size_t i = 0; i < haystack_size; ++i)
{
const bool match = regexp.getRE2()->Match(
{reinterpret_cast<const char *>(&haystack[offset]), N},
0,
N,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match;
offset += N;
}
}
}
else
{
/// NOTE This almost matches with the case of likePatternIsSubstring.
const UInt8 * const begin = haystack.data();
const UInt8 * const end = haystack.data() + haystack.size();
const UInt8 * pos = begin;
size_t i = 0;
const UInt8 * next_pos = begin;
/// If required substring is larger than string size - it cannot be found.
if (required_substring.size() <= N)
{
Searcher searcher(required_substring.data(), required_substring.size(), end - pos);
/// We will search for the next occurrence in all rows at once.
while (pos < end && end != (pos = searcher.search(pos, end - pos)))
{
/// Let's determine which index it refers to.
while (next_pos + N <= pos)
{
res[i] = negate;
next_pos += N;
++i;
}
next_pos += N;
if (pos + required_substring.size() <= next_pos)
{
/// And if it does not, if necessary, we check the regexp.
if (is_trivial)
res[i] = !negate;
else
{
const char * str_data = reinterpret_cast<const char *>(next_pos - N);
/** Even in the case of `required_substring_is_prefix` use UNANCHORED check for regexp,
* so that it can match when `required_substring` occurs into the string several times,
* and at the first occurrence, the regexp is not a match.
*/
const size_t start_pos = (required_substring_is_prefix) ? (reinterpret_cast<const char *>(pos) - str_data) : 0;
const size_t end_pos = N;
const bool match = regexp.getRE2()->Match(
{str_data, N},
start_pos,
end_pos,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match;
}
}
else
res[i] = negate;
pos = next_pos;
++i;
}
}
/// Tail, in which there can be no substring.
if (i < res.size())
memset(&res[i], negate, (res.size() - i) * sizeof(res[0]));
}
}
static void vectorVector(
const ColumnString::Chars & haystack_data,
const ColumnString::Offsets & haystack_offsets,
const ColumnString::Chars & needle_data,
const ColumnString::Offsets & needle_offset,
[[maybe_unused]] const ColumnPtr & start_pos_,
PaddedPODArray<UInt8> & res,
[[maybe_unused]] ColumnUInt8 * res_null)
{
/// `res_null` serves as an output parameter for implementing an XYZOrNull variant.
assert(!res_null);
const size_t haystack_size = haystack_offsets.size();
assert(haystack_size == needle_offset.size());
assert(haystack_size == res.size());
assert(start_pos_ == nullptr);
if (haystack_offsets.empty())
return;
String required_substr;
bool is_trivial;
bool required_substring_is_prefix; /// for `anchored` execution of the regexp.
size_t prev_haystack_offset = 0;
size_t prev_needle_offset = 0;
Regexps::LocalCacheTable cache;
Regexps::RegexpPtr regexp;
for (size_t i = 0; i < haystack_size; ++i)
{
const auto * const cur_haystack_data = &haystack_data[prev_haystack_offset];
const size_t cur_haystack_length = haystack_offsets[i] - prev_haystack_offset - 1;
const auto * const cur_needle_data = &needle_data[prev_needle_offset];
const size_t cur_needle_length = needle_offset[i] - prev_needle_offset - 1;
const auto & needle = String(
reinterpret_cast<const char *>(cur_needle_data),
cur_needle_length);
if (is_like && impl::likePatternIsSubstring(needle, required_substr))
{
if (required_substr.size() > cur_haystack_length)
res[i] = negate;
else
{
Searcher searcher(required_substr.data(), required_substr.size(), cur_haystack_length);
const auto * match = searcher.search(cur_haystack_data, cur_haystack_length);
res[i] = negate ^ (match != cur_haystack_data + cur_haystack_length);
}
}
else
{
regexp = cache.getOrSet<is_like, /*no_capture*/ true, case_insensitive>(needle);
regexp->getAnalyzeResult(required_substr, is_trivial, required_substring_is_prefix);
if (required_substr.empty())
{
if (!regexp->getRE2()) /// An empty regexp. Always matches.
res[i] = !negate;
else
{
const bool match = regexp->getRE2()->Match(
{reinterpret_cast<const char *>(cur_haystack_data), cur_haystack_length},
0,
cur_haystack_length,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match;
}
}
else
{
Searcher searcher(required_substr.data(), required_substr.size(), cur_haystack_length);
const auto * match = searcher.search(cur_haystack_data, cur_haystack_length);
if (match == cur_haystack_data + cur_haystack_length)
res[i] = negate; // no match
else
{
if (is_trivial)
res[i] = !negate; // no wildcards in pattern
else
{
const size_t start_pos = (required_substring_is_prefix) ? (match - cur_haystack_data) : 0;
const size_t end_pos = cur_haystack_length;
const bool match2 = regexp->getRE2()->Match(
{reinterpret_cast<const char *>(cur_haystack_data), cur_haystack_length},
start_pos,
end_pos,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match2;
}
}
}
}
prev_haystack_offset = haystack_offsets[i];
prev_needle_offset = needle_offset[i];
}
}
static void vectorFixedVector(
const ColumnString::Chars & haystack,
size_t N,
const ColumnString::Chars & needle_data,
const ColumnString::Offsets & needle_offset,
[[maybe_unused]] const ColumnPtr & start_pos_,
PaddedPODArray<UInt8> & res,
[[maybe_unused]] ColumnUInt8 * res_null)
{
/// `res_null` serves as an output parameter for implementing an XYZOrNull variant.
assert(!res_null);
const size_t haystack_size = haystack.size()/N;
assert(haystack_size == needle_offset.size());
assert(haystack_size == res.size());
assert(start_pos_ == nullptr);
if (haystack.empty())
return;
String required_substr;
bool is_trivial;
bool required_substring_is_prefix; // for `anchored` execution of the regexp.
size_t prev_haystack_offset = 0;
size_t prev_needle_offset = 0;
Regexps::LocalCacheTable cache;
Regexps::RegexpPtr regexp;
for (size_t i = 0; i < haystack_size; ++i)
{
const auto * const cur_haystack_data = &haystack[prev_haystack_offset];
const size_t cur_haystack_length = N;
const auto * const cur_needle_data = &needle_data[prev_needle_offset];
const size_t cur_needle_length = needle_offset[i] - prev_needle_offset - 1;
const auto & needle = String(
reinterpret_cast<const char *>(cur_needle_data),
cur_needle_length);
if (is_like && impl::likePatternIsSubstring(needle, required_substr))
{
if (required_substr.size() > cur_haystack_length)
res[i] = negate;
else
{
Searcher searcher(required_substr.data(), required_substr.size(), cur_haystack_length);
const auto * match = searcher.search(cur_haystack_data, cur_haystack_length);
res[i] = negate ^ (match != cur_haystack_data + cur_haystack_length);
}
}
else
{
regexp = cache.getOrSet<is_like, /*no_capture*/ true, case_insensitive>(needle);
regexp->getAnalyzeResult(required_substr, is_trivial, required_substring_is_prefix);
if (required_substr.empty())
{
if (!regexp->getRE2()) /// An empty regexp. Always matches.
res[i] = !negate;
else
{
const bool match = regexp->getRE2()->Match(
{reinterpret_cast<const char *>(cur_haystack_data), cur_haystack_length},
0,
cur_haystack_length,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match;
}
}
else
{
Searcher searcher(required_substr.data(), required_substr.size(), cur_haystack_length);
const auto * match = searcher.search(cur_haystack_data, cur_haystack_length);
if (match == cur_haystack_data + cur_haystack_length)
res[i] = negate; // no match
else
{
if (is_trivial)
res[i] = !negate; // no wildcards in pattern
else
{
const size_t start_pos = (required_substring_is_prefix) ? (match - cur_haystack_data) : 0;
const size_t end_pos = cur_haystack_length;
const bool match2 = regexp->getRE2()->Match(
{reinterpret_cast<const char *>(cur_haystack_data), cur_haystack_length},
start_pos,
end_pos,
re2::RE2::UNANCHORED,
nullptr,
0);
res[i] = negate ^ match2;
}
}
}
}
prev_haystack_offset += N;
prev_needle_offset = needle_offset[i];
}
}
template <typename... Args>
static void constantVector(Args &&...)
{
throw Exception(ErrorCodes::ILLEGAL_COLUMN, "Function '{}' doesn't support search with non-constant needles in constant haystack", name);
}
};
}