WIP: replaced diff to diff-match-patch (#3675)

[jalnanco]

Description of the Change

I checked the #3675 issue and changed the existing diff library to Google's diff-patch-match.

Benefits

• Diff speed is improved for large data.

Test Code (copy from issue)

var _ = require('lodash');
const {expect} = require('chai')
it('Test failing', function () {
  const longArray = _.times(10000, function (i) {
    return {a : i}
  })
  const shortArray = []
  expect(longArray).deep.equal(shortArray)
}

Before change: About 2 minutes 1 second

/bin/mocha ../../pycon/mocha/my/test_issue  121.94s user 0.55s system 100% cpu 2:01.74 total

After change: About 1 second

./bin/mocha ../../pycon/mocha/my/test_issue  0.91s user 0.09s system 93% cpu 1.071 total

Applicable issues

Diff-patch-match is character-based, unlike diff.
So it is written to be processed line by line using line-mode.

However, the result screen is different because there is a difference from Patch Process.

Test Code

it('Test failing2', function () {
  const expected = { a: 123 };
  const actual = { a: 123, b: 456 };
  expect(expected).deep.equal(actual);
})

Comparison of Patch Results

Before

Index: string
===================================================================
--- string
+++ string
@@ -1,3 +1,4 @@
 {
   "a": 123
+  "b": 456
 }
\ No newline at end of file

After

@@ -6,9 +6,20 @@
 a": 123

+  "b": 456

 }

Comparison of Excution Results

Before

       {
         "a": 123
      +  "b": 456
       }

After

       a": 123
      +  "b": 456
       }

RESULT

So I don't think it's user friendly.

Would it be better to work on the same output as the existing result? Or would you like to hear other ways to find out?

Please, I want to listen to the comments and advices.

• All test code is run on MacBook Pro 2017 (16GB)
• reference links
  Line mode
  That library is character-based

[jsf-clabot]

All committers have signed the CLA.

[wnghdcjfe]

I tested it with your code and there was no problem. What part was wrong?

const {expect} = require('chai')
it('Test failing2', function () {
    const expected = { a: 123 };
    const actual = { a: 123, b: 456 };
    expect(expected).deep.equal(actual);
  })

  1) Test failing2:

      AssertionError: expected { a: 123 } to deeply equal { a: 123, b: 456 }
      + expected - actual

       {
         "a": 123
      +  "b": 456
       }

[jalnanco]

I tested it with your code and there was no problem. What part was wrong?

const {expect} = require('chai')
it('Test failing2', function () {
    const expected = { a: 123 };
    const actual = { a: 123, b: 456 };
    expect(expected).deep.equal(actual);
  })

  1) Test failing2:

      AssertionError: expected { a: 123 } to deeply equal { a: 123, b: 456 }
      + expected - actual

       {
         "a": 123
      +  "b": 456
       }

If you use Google's diff-patch-match instead of old diff.
You can get speed but It's difficult to recognize.
Because It's consist of character-based.

So Test is not a problem. Result output is a problem.