gaspra 0.1.0a3

A fast Python tool for searching, diffing, and merging text

What is GASPRA

GASPRA is the Generalized Automaton-based String Processing for Rapid Alignment. GASPRA implements an efficient algorithm to solve the Longest Common Substring (LCS) problem. Specifically it uses a suffix automaton, which is a finite state machine and is extremely efficient at string searching and matching problems. It uses the suffix automaton as a base on which to build diffing and merging tools. Because it's efficient, it's able to apply correct merge files viewing them as sequences of characters, where most traditional merge programs view files as sequences of lines. Documents with a lot of multi-line text (e.g., markdown files, and LaTex files, and lengthy comments in source code) confuse line-oriented merge algorithms, and you end up with a lot of "conflicts" that don't really exist.

The line-oriented approach is an order of magnitude faster, and GASPRA allows you to choose line-oriented or character-oriented merges. Both will use the suffix automaton approach. It's a matter of whether you view a text file as a sequence of characters or a sequence of lines. It's still a sequence of tokens and the suffix automaton approach works on both cases.

Here's an example of running gaspra-merge on three files were the first had no line breaks, the second was edited (without adding line breaks), and the third was reflowed. These files were merged with no conflicts. Line-oriented tools show the entire paragraph as a conflict. The result shown is the output of the command:

gaspra-merge -sdc data/tale0 data/tale1-reflowed data/tale2-edited

See Merging for another example.

What's included

• gaspra-diff - A tool for diffing two text files in either a line-oriented or a character-oriented fashion.
• gaspra-merge - A tool for doing a 3-way merge in either a line-oriented or a character-oriented fashion.

GASPRA has been tested on Bill Ritcher's pathological merge test casea and performs quite well

Why GASPRA exists

Recently, I became interested in string matching. In particular, I was interested in the LCS problem and related problems. I discovered that efficient string matching is hard to find, especially in Python. Much of what you find on the internet is either slow, or in some cases, simply doesn't work. I wanted a pure Python solution that is reasonably fast to experiment with some string matching algorithms.

Though implementations appear uncommon, efficient algorithms exist. The LCS problem is solvable in linear time, yet most online examples use a classical dynamic programming approach which is quadratic in both space and time. Even the venerable difflib, which is part of Python's standard library uses quadratic-time algorithms. GASPRA uses efficient suffix automata with linear time and space complexity. The difference is dramatic. It's the same as the difference between a quick-sort and a bubble sort. Because of this difference, nobody uses a bubble sort, which is universally recognized as a naive approach. Yet somehow, using dynamic programming to solve text matching is widely accepted. Difflib is practically unusable on large, document-length strings. The table below shows the time to find the LCS in two strings for difflib compared to this package. The strings are equal-length, random sequences of the letters a, b, and c. The lengths shown in the table are the combined length of the two strings. The quadratic complexity of 'difflib' is obvious. Even though it's a pure Python solution, GASPRA is quite fast by comparison.

Length	Match Length	Difflib (ms)	GASPRA (ms)
2k	13	31	3
4k	13	120	5
8k	15	472	11
16k	15	1,897	32
32k	16	7,828	52
64k	18	33,314	121
128k	19	135,413	341
256k	23	554,665	690

Installation

You can install GASPRA using pip (or similar compatible tools):

pip install gaspra

Examples.

Finding Longest Common Substrings

The function gaspra.find_lcs() returns the starting positions and the length of common substring of two strings with maximal length:

>>> import gaspra                                                                                               
>>> a="Call me Ishmael. Some years ago—never mind how long \
precisely—having little or no money in my purse, and nothing \
particular to interest me on shore, I thought I would sail \
about a little and see the watery part of the world"                                                                                                                  
>>> b="It was the best of times, it was the worst of times, \
it was the age of wisdom, it was the age of foolishness, \
it was the epoch of belief, it was the epoch of incredulity, \
it was the season of Light, it was the season of Darkness, \
it was the spring of hope, it was the winter of despair, we \
had everything before us, we had nothing before us, we were \
all going direct to Heaven, we were all going direct the other \
way – in short, the period was so far like the present period, \
that some of its noisiest authorities insisted on its being \
received, for good or for evil, in the superlative degree of \
comparison only."
>>> gaspra.find_lcs(a,b)
(103, 321, 10)
>>> a[103:103+10]
'd nothing '
>>> b[321:321+10]
'd nothing '

There's also find_lcs_multiple() for n-way common strings:

>>> a="The quick brown fox jumps over the lazy dog near the riverbank."
>>> b="The quick brown fox leaps over the lazy dogs near the river."
>>> c="The quick, clever fox jumps across the lazy dogs by the riverbank."
>>> gaspra.find_lcs_multiple(a, b, c)
((30, 30, 34), 13)
>>> a[30:30+13]
' the lazy dog'

Finding Diffs

Given an original string and a modified string, the function gaspra.diff() returns the sequence of changes represented interspersed with fragments from the original string. The sequence is a sequence of strings (fragments from the original) and tuples of two strings representing an insertion/deletion pair. Note that an insertion is a tuple where the deletion string is empty, and vice versa.

>>> import gaspra
>>> original="The quick brown fox jumps over the lazy dog near the riverbank."
>>> modified="The quick brown fox leaps over the lazy dogs near the river"
>>> list(gaspra.diff(original, modified))
['The quick brown fox ', ('lea', 'jum'), 'ps over the lazy dog', ('s', ''), ' near the river', ('', 'bank.')]

Merging

Here's an example of two different revisions of the same sentence that can be resolved without any conflicts.

>>> original = "The quick brown fox jumps over the lazy dog near the riverbank."
>>> editor1 = "The quick brown fox leaps over the lazy dogs near the river."
>>> editor2 = "The quick, clever fox jumps across the lazy dogs by the riverbank."
>>> list(gaspra.diff(original, editor1))
['The quick brown fox ', ('lea', 'jum'), 'ps over the lazy dog', ('s', ''), ' near the river', ('', 'bank'), '.']
>>> list(gaspra.diff(original, editor2))
['The quick', (',', ''), ' ', ('cleve', 'b'), 'r', ('', 'own'), ' fox jumps ', ('ac', 'ove'), 'r', ('oss', ''), ' the lazy dog', ('s', ''), ' ', ('by', 'near'), ' the riverbank.']
>>> list(gaspra.merge(original, editor1, editor2))
['The quick, clever fox leaps across the lazy dogs by the river.']

Here's the same example but with a different revision by editor2 that does conflict with editor1. The conflicts are represented as a tuple with two alternate versions of the merged text.

>>> conflicts_with_1 = "The swift, agile fox leaps over the sleepy dog near the riverside."
>>> list(gaspra.diff(original, conflicts_with_1))
['The ', ('s', 'quick bro'), 'w', ('ift, agile', 'n'), ' fox ', ('lea', 'jum'), 'ps over the ', ('s', ''), 'l', ('eep', 'az'), 'y dog near the river', ('side', 'bank'), '.']
>>> list(gaspra.merge(original, editor1, conflicts_with_1))
['The swift, agile fox leaps over the sleepy dogs near the river', ('', 'side'), '.']