Compare items in two sequences to find a longest common subsequence.
The items not in common are the items to delete or insert in a shortest edit script.
To maximize flexibility and minimize memory, you write callback functions as configuration:
Input function isCommon(aIndex, bIndex) compares items at indexes in the sequences and returns a truthy/falsey value. This package might call your function more than once for some pairs of indexes.
Because your function encapsulates comparison, this package can compare items according to === operator, Object.is method, or other criterion.
Because your function encapsulates sequences, this package can find differences in arrays, strings, or other data.
Output function foundSubsequence(nCommon, aCommon, bCommon) receives the number of adjacent items and starting indexes of each common subsequence. If sequences do not have common items, then this package does not call your function.
If N is the sum of lengths of sequences and L is length of a longest common subsequence, then D = N – 2L is the number of differences in the corresponding shortest edit script.
Some sequences (for example, a and b in the example of usage) have more than one longest common subsequence.
This package finds the following common items:
comparisons of common items
values
output arguments
a[2] === b[0]
'c'
foundSubsequence(1, 2, 0)
a[4] === b[1]
'b'
foundSubsequence(1, 4, 1)
a[5] === b[3] && a[6] === b[4]
'b', 'a'
foundSubsequence(2, 5, 3)
The “edit graph” analogy in the Myers paper shows the following common items:
comparisons of common items
values
a[2] === b[0]
'c'
a[3] === b[2] && a[4] === b[3]
'a', 'b'
a[6] === b[4]
'a'
Various packages which implement the Myers algorithm will always agree on the length of a longest common subsequence, but might sometimes disagree on which items are in it.
Example of callback functions to count common items
// Return length of longest common subsequence according to === operator.function countCommonItems(a, b) { let n = 0; function isCommon(aIndex, bIndex) { return a[aIndex] === b[bIndex]; } function foundSubsequence(nCommon) { n += nCommon; } diff(a.length, b.length, isCommon, foundSubsequence); return n;}const commonLength = countCommonItems( ['a', 'b', 'c', 'a', 'b', 'b', 'a'], ['c'
category of items
expression
value
in common
commonLength
4
to delete from a
a.length - commonLength
3
to insert from b
b.length - commonLength
2
If the length difference b.length - a.length is:
negative: its absolute value is the minimum number of items to delete from a
positive: it is the minimum number of items to insert from b
zero: there is an equal number of items to delete from a and insert from b
non-zero: there is an equal number of additional items to delete from a and insert from b
In this example, 6 - 7 is:
negative: 1 is the minimum number of items to delete from a
non-zero: 2 is the number of additional items to delete from a and insert from b
Example of callback functions to find common items
// Return array of items in longest common subsequence according to Object.is method.const findCommonItems = (a, b) => { const array = []; diff( a.length, b.length, (aIndex, bIndex) => Object.is(a[aIndex], b[bIndex]), (nCommon, aCommon) => { for (; nCommon !== 0; nCommon -= 1, aCommon += 1) { array.push(a[aCommon]); } }, ); return array;};const commonItems = findCommonItems( ['a',
i
commonItems[i]
aIndex
0
'c'
2
1
'b'
4
2
'b'
5
3
'a'
6
Example of callback functions to diff index intervals
Instead of slicing array-like objects, you can adjust indexes in your callback functions.
// Diff index intervals that are half open [start, end) like array slice method.const diffIndexIntervals = (a, aStart, aEnd, b, bStart, bEnd) => { // Validate: 0 <= aStart and aStart <= aEnd and aEnd <= a.length // Validate: 0 <= bStart and bStart <= bEnd and bEnd <= b.length diff( aEnd - aStart, bEnd - bStart, (aIndex, bIndex) => Object.is(a[aStart + aIndex], b[bStart + bIndex]), (nCommon, aCommon, bCommon) => { // aStart + aCommon, bStart + bCommon }, ); // After the last common subsequence, do any remaining work.};
Example of callback functions to emulate diff command
Linux or Unix has a diff command to compare files line by line. Its output is a shortest edit script:
change adjacent lines from the first file to lines from the second file
delete lines from the first file
append or insert lines from the second file
// Given zero-based half-open range [start, end) of array indexes,// return one-based closed range [start + 1, end] as string.const getRange = (start, end) => start + 1 === end ? `${start + 1}` : `${start + 1},${end}`;// Given index intervals of lines to delete or insert, or both, or neither,// push formatted diff lines onto array.const pushDelIns = (aLines, aIndex, aEnd, bLines, bIndex, bEnd, array) => { const deleteLines = aIndex !== aEnd; const insertLines = bIndex !== bEnd; const changeLines = deleteLines && insertLines;
Example of callback functions to format diff lines
Here is simplified code to format changed and unchanged lines in expected and received values after a test fails in Jest:
// Format diff with minus or plus for change lines and space for common lines.const formatDiffLines = (a, b) => { // Jest depends on pretty-format package to serialize objects as strings. // Unindented for comparison to avoid distracting differences: const aLinesUn = format(a, {indent: 0 /*, other options*/}).split('\n'); const bLinesUn = format(b, {indent: 0 /*, other options*/}).split('\n'); // Indented to display changed and unchanged lines: const aLinesIn = format(a, {indent: 2 /*, other options*/}).split('\n'); const bLinesIn =
If N is the sum of lengths of sequences and L is length of a longest common subsequence, then N – L is length of an array of diff lines. In this example, N is 7 + 9, L is 5, and N – L is 11.
i
diffLines[i]
aIndex
bIndex
0
'··Object {'
0
0
1
'····"searching": "",'
1
1
2
'-···"sorting": Object {'
2
3
'-·····"ascending": true,'
3
4
'+·····"sorting": Array ['
Example of callback functions to find diff items
Here is simplified code to find changed and unchanged substrings within adjacent changed lines in expected and received values after a test fails in Jest:
// Return diff items for strings (compatible with diff-match-patch package).const findDiffItems = (a, b) => { const isCommon = (aIndex, bIndex) => a[aIndex] === b[bIndex]; let aIndex = 0; let bIndex = 0; const array = []; const foundSubsequence = (nCommon, aCommon, bCommon) => { if (aIndex !== aCommon) { array.push([-1, a.slice(aIndex, aCommon)]); // delete is -1 } if (bIndex !== bCommon) { array.push
i
diffItems[i][0]
diffItems[i][1]
0
0
'"sorting": '
1
1
'Array [\n'
2
0
'Object {\n"'
3
-1
'a'
4
1
'de'
5
0
'scending": '
6
The length difference b.length - a.length is equal to the sum of diffItems[i][0] values times diffItems[i][1] lengths. In this example, the difference 48 - 38 is equal to the sum 10.
category of diff item
[0]
[1] lengths
subtotal
in common
0
11 + 10 + 11 + 2
0
to delete from a
–1
1 + 3
-4
to insert from b
1
8 + 2 + 4
14
Instead of formatting the changed substrings with escape codes for colors in the foundSubsequence function to save memory, this example spends memory to gain flexibility before formatting, so a separate heuristic algorithm might modify the generic array of diff items to show changes more clearly:
i
diffItems[i][0]
diffItems[i][1]
6
-1
'true'
7
1
'false'
8
0
','
For expected and received strings of serialized data, the result of finding changed lines, and then finding changed substrings within adjacent changed lines (as in the preceding two examples) sometimes displays the changes in a more intuitive way than the result of finding changed substrings, and then splitting them into changed and unchanged lines.