Functions for [replacing](../../sql-reference/functions/string-replace-functions.md) and [other manipulations with strings](../../sql-reference/functions/string-functions.md) are described separately.
Works under the assumption that the string contains a set of bytes representing a single-byte encoded text. If this assumption is not met and a character can’t be represented using a single byte, the function doesn’t throw an exception and returns some unexpected result. If character can be represented using two bytes, it will use two bytes and so on.
The same phrase in Russian contains characters which can’t be represented using a single byte. The function returns some unexpected result (use [positionUTF8](#positionutf8) function for multi-byte encoded text):
The same as [position](#position) returns the position (in bytes) of the found substring in the string, starting from 1. Use the function for a case-insensitive search.
Works under the assumption that the string contains a set of bytes representing a single-byte encoded text. If this assumption is not met and a character can’t be represented using a single byte, the function doesn’t throw an exception and returns some unexpected result. If character can be represented using two bytes, it will use two bytes and so on.
Works under the assumption that the string contains a set of bytes representing a UTF-8 encoded text. If this assumption is not met, the function doesn’t throw an exception and returns some unexpected result. If character can be represented using two Unicode points, it will use two and so on.
The phrase “Hello, world!” in Russian contains a set of Unicode points representing a single-point encoded text. The function returns some expected result:
The phrase “Salut, étudiante!”, where character `é` can be represented using a one point (`U+00E9`) or two points (`U+0065U+0301`) the function can be returned some unexpected result:
The same as [positionUTF8](#positionutf8), but is case-insensitive. Returns the position (in Unicode points) of the found substring in the string, starting from 1.
Works under the assumption that the string contains a set of bytes representing a UTF-8 encoded text. If this assumption is not met, the function doesn’t throw an exception and returns some unexpected result. If character can be represented using two Unicode points, it will use two and so on.
The same as [position](../../sql-reference/functions/string-search-functions.md#position) but returns `Array` of positions (in bytes) of the found corresponding substrings in the string. Positions are indexed starting from 1.
For a case-insensitive search or/and in UTF-8 format use functions `multiSearchFirstPositionCaseInsensitive, multiSearchFirstPositionUTF8, multiSearchFirstPositionCaseInsensitiveUTF8`.
For a case-insensitive search or/and in UTF-8 format use functions `multiSearchFirstIndexCaseInsensitive, multiSearchFirstIndexUTF8, multiSearchFirstIndexCaseInsensitiveUTF8`.
For a case-insensitive search or/and in UTF-8 format use functions `multiSearchAnyCaseInsensitive, multiSearchAnyUTF8, multiSearchAnyCaseInsensitiveUTF8`.
Checks whether the string matches the `pattern` regular expression. A `re2` regular expression. The [syntax](https://github.com/google/re2/wiki/Syntax) of the `re2` regular expressions is more limited than the syntax of the Perl regular expressions.
Note that the backslash symbol (`\`) is used for escaping in the regular expression. The same symbol is used for escaping in string literals. So in order to escape the symbol in a regular expression, you must write two backslashes (\\) in a string literal.
The same as `match`, but returns 0 if none of the regular expressions are matched and 1 if any of the patterns matches. It uses [hyperscan](https://github.com/intel/hyperscan) library. For patterns to search substrings in a string, it is better to use `multiSearchAny` since it works much faster.
The length of any of the `haystack` string must be less than 2<sup>32</sup> bytes otherwise the exception is thrown. This restriction takes place because of hyperscan API.
The same as `multiMatchAny`, but returns 1 if any pattern matches the haystack within a constant [edit distance](https://en.wikipedia.org/wiki/Edit_distance). This function is also in an experimental mode and can be extremely slow. For more information see [hyperscan documentation](https://intel.github.io/hyperscan/dev-reference/compilation.html#approximate-matching).
Extracts a fragment of a string using a regular expression. If ‘haystack’ doesn’t match the ‘pattern’ regex, an empty string is returned. If the regex doesn’t contain subpatterns, it takes the fragment that matches the entire regex. Otherwise, it takes the fragment that matches the first subpattern.
Extracts all the fragments of a string using a regular expression. If ‘haystack’ doesn’t match the ‘pattern’ regex, an empty string is returned. Returns an array of strings consisting of all matches to the regex. In general, the behavior is the same as the ‘extract’ function (it takes the first subpattern, or the entire expression if there isn’t a subpattern).
Calculates the 4-gram distance between `haystack` and `needle`: counts the symmetric difference between two multisets of 4-grams and normalizes it by the sum of their cardinalities. Returns float number from 0 to 1 – the closer to zero, the more strings are similar to each other. If the constant `needle` or `haystack` is more than 32Kb, throws an exception. If some of the non-constant `haystack` or `needle` strings are more than 32Kb, the distance is always one.
Same as `ngramDistance` but calculates the non-symmetric difference between `needle` and `haystack`– the number of n-grams from needle minus the common number of n-grams normalized by the number of `needle` n-grams. The closer to one, the more likely `needle` is in the `haystack`. Can be useful for fuzzy string search.
For UTF-8 case we use 3-gram distance. All these are not perfectly fair n-gram distances. We use 2-byte hashes to hash n-grams and then calculate the (non-)symmetric difference between these hash tables – collisions may occur. With UTF-8 case-insensitive format we do not use fair `tolower` function – we zero the 5-th bit (starting from zero) of each codepoint byte and first bit of zeroth byte if bytes more than one – this works for Latin and mostly for all Cyrillic letters.
