|
Data.ByteString.Lazy.Search | Portability | non-portable (BangPatterns) | Stability | Provisional | Maintainer | Daniel Fischer <daniel.is.fischer@web.de> |
|
|
|
|
|
Description |
Fast overlapping Boyer-Moore search of lazy
ByteString values. Breaking, splitting and replacing
using the Boyer-Moore algorithm.
Descriptions of the algorithm can be found at
http://www-igm.univ-mlv.fr/~lecroq/string/node14.html#SECTION00140
and
http://en.wikipedia.org/wiki/Boyer-Moore_string_search_algorithm
Original authors: Daniel Fischer (daniel.is.fischer at web.de) and
Chris Kuklewicz (haskell at list.mightyreason.com).
|
|
Synopsis |
|
|
|
|
Overview
|
|
This module provides functions related to searching a substring within
a string, using the Boyer-Moore algorithm with minor modifications
to improve the overall performance and ameliorate the worst case
performance degradation of the original Boyer-Moore algorithm for
periodic patterns.
Efficiency demands that the pattern be a strict ByteString,
to work with a lazy pattern, convert it to a strict ByteString
first via strictify (provided it is not too long).
If support for long lazy patterns is needed, mail a feature-request.
When searching a pattern in a UTF-8-encoded ByteString, be aware that
these functions work on bytes, not characters, so the indices are
byte-offsets, not character offsets.
|
|
Performance
|
|
In general, the Boyer-Moore algorithm is the most efficient method to
search for a pattern inside a string. The advantage over other algorithms
(e.g. Naïve, Knuth-Morris-Pratt, Horspool, Sunday) can be made
arbitrarily large for specially selected patterns and targets, but
usually, it's a factor of 2–3 versus Knuth-Morris-Pratt and of
6–10 versus the naïve algorithm. The Horspool and Sunday
algorithms, which are simplified variants of the Boyer-Moore algorithm,
typically have performance between Boyer-Moore and Knuth-Morris-Pratt,
mostly closer to Boyer-Moore. The advantage of the Boyer-moore variants
over other algorithms generally becomes larger for longer patterns. For
very short patterns (or patterns with a very short period), other
algorithms, e.g. Data.ByteString.Lazy.Search.DFA can be faster (my
tests suggest that "very short" means two, maybe three bytes).
In general, searching in a strict ByteString is slightly faster
than searching in a lazy ByteString, but for long targets the
smaller memory footprint of lazy L.ByteStrings can make searching
those (sometimes much) faster. On the other hand, there are cases
where searching in a strict target is much faster, even for long targets.
On 32-bit systems, Int-arithmetic is much faster than Int64-arithmetic,
so when there are many matches, that can make a significant difference.
Also, the modification to ameliorate the case of periodic patterns
is defeated by chunk-boundaries, so long patterns with a short period
and many matches exhibit poor behaviour (consider using indices from
Data.ByteString.Lazy.Search.DFA or Data.ByteString.Lazy.Search.KMP
in those cases, the former for medium-length patterns, the latter for
long patterns; none of the functions except indices suffer from
this problem, though).
|
|
Caution
|
|
When working with a lazy target string, the relation between the pattern
length and the chunk size can play a big rôle.
Crossing chunk boundaries is relatively expensive, so when that becomes
a frequent occurrence, as may happen when the pattern length is close
to or larger than the chunk size, performance is likely to degrade.
If it is needed, steps can be taken to ameliorate that effect, but unless
entirely separate functions are introduced, that would hurt the
performance for the more common case of patterns much shorter than
the default chunk size.
|
|
Complexity
|
|
Preprocessing the pattern is O(patternLength + σ) in time and
space (σ is the alphabet size, 256 here) for all functions.
The time complexity of the searching phase for indices
is O(targetLength / patternLength) in the best case.
For non-periodic patterns, the worst case complexity is
O(targetLength), but for periodic patterns, the worst case complexity
is O(targetLength * patternLength) for the original Boyer-Moore
algorithm.
The searching functions in this module contain a modification which
drastically improves the performance for periodic patterns, although
less for lazy targets than for strict ones.
If I'm not mistaken, the worst case complexity for periodic patterns
is O(targetLength * (1 + patternLength / chunkSize)).
The other functions don't have to deal with possible overlapping
patterns, hence the worst case complexity for the processing phase
is O(targetLength) (respectively O(firstIndex + patternLength)
for the breaking functions if the pattern occurs).
|
|
Partial application
|
|
All functions can usefully be partially applied. Given only a pattern,
the pattern is preprocessed only once, allowing efficient re-use.
|
|
Integer overflow
|
|
The current code uses Int to keep track of the locations in the
target string. If the length of the pattern plus the length of any
strict chunk of the target string is greater or equal to
maxBound :: Int then this will overflow causing an error. We try
to detect this and call error before a segfault occurs.
|
|
Finding substrings
|
|
|
:: ByteString | Strict pattern to find
| -> ByteString | Lazy string to search
| -> [Int64] | Offsets of matches
| indices finds the starting indices of all possibly overlapping
occurrences of the pattern in the target string.
If the pattern is empty, the result is [0 .. length target].
|
|
|
|
:: ByteString | Strict pattern to find
| -> ByteString | Lazy string to search
| -> [Int64] | Offsets of matches
| nonOverlappingIndices finds the starting indices of all
non-overlapping occurrences of the pattern in the target string.
It is more efficient than removing indices from the list produced
by indices.
|
|
|
Breaking on substrings
|
|
|
:: ByteString | Strict pattern to search for
| -> ByteString | Lazy string to search in
| -> (ByteString, ByteString) | Head and tail of string broken at substring
| breakOn pattern target splits target at the first occurrence
of pattern. If the pattern does not occur in the target, the
second component of the result is empty, otherwise it starts with
pattern. If the pattern is empty, the first component is empty.
For a non-empty pattern, the first component is generated lazily,
thus the first parts of it can be available before the pattern has
been found or determined to be absent.
uncurry append . breakOn pattern = id
|
|
|
|
:: ByteString | Strict pattern to search for
| -> ByteString | Lazy string to search in
| -> (ByteString, ByteString) | Head and tail of string broken after substring
| breakAfter pattern target splits target behind the first occurrence
of pattern. An empty second component means that either the pattern
does not occur in the target or the first occurrence of pattern is at
the very end of target. If you need to discriminate between those cases,
use breakFindAfter.
If the pattern is empty, the first component is empty.
For a non-empty pattern, the first component is generated lazily,
thus the first parts of it can be available before the pattern has
been found or determined to be absent.
uncurry append . breakAfter pattern = id
|
|
|
|
|
|
Replacing
|
|
|
:: Substitution rep | | => ByteString | Strict pattern to replace
| -> rep | Replacement string
| -> ByteString | Lazy string to modify
| -> ByteString | Lazy result
| replace pat sub text replaces all (non-overlapping) occurrences of
pat in text with sub. If occurrences of pat overlap, the first
occurrence that does not overlap with a replaced previous occurrence
is substituted. Occurrences of pat arising from a substitution
will not be substituted. For example:
replace "ana" "olog" "banana" = "bologna"
replace "ana" "o" "bananana" = "bono"
replace "aab" "abaa" "aaab" = "abaaab"
The result is a lazy ByteString,
which is lazily produced, without copying.
Equality of pattern and substitution is not checked, but
replace pat pat text == text
holds (the internal structure is generally different).
If the pattern is empty but not the substitution, the result
is equivalent to (were they Strings) cycle sub.
For non-empty pat and sub a lazy ByteString,
concat . Data.List.intersperse sub . split pat = replace pat sub
and analogous relations hold for other types of sub.
|
|
|
Splitting
|
|
|
:: ByteString | Strict pattern to split on
| -> ByteString | Lazy string to split
| -> [ByteString] | Fragments of string
| split pattern target splits target at each (non-overlapping)
occurrence of pattern, removing pattern. If pattern is empty,
the result is an infinite list of empty ByteStrings, if target
is empty but not pattern, the result is an empty list, otherwise
the following relations hold (where patL is the lazy ByteString
corresponding to pat):
concat . Data.List.intersperse patL . split pat = id,
length (split pattern target) ==
length (nonOverlappingIndices pattern target) + 1,
no fragment in the result contains an occurrence of pattern.
|
|
|
|
:: ByteString | Strict pattern to split on
| -> ByteString | Lazy string to split
| -> [ByteString] | Fragments of string
| splitKeepEnd pattern target splits target after each (non-overlapping)
occurrence of pattern. If pattern is empty, the result is an
infinite list of empty ByteStrings, otherwise the following
relations hold:
concat . splitKeepEnd pattern = id,
all fragments in the result except possibly the last end with
pattern, no fragment contains more than one occurrence of pattern.
|
|
|
|
:: ByteString | Strict pattern to split on
| -> ByteString | Lazy string to split
| -> [ByteString] | Fragments of string
| splitKeepFront is like splitKeepEnd, except that target is split
before each occurrence of pattern and hence all fragments
with the possible exception of the first begin with pattern.
|
|
|
Convenience
|
|
|
strictify converts a lazy ByteString to a strict ByteString
to make it a suitable pattern.
|
|
Produced by Haddock version 2.4.2 |