Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Incorrect results when Enumerable contains generic objects. String collections works. #35

Closed
tomasjura opened this issue Dec 3, 2015 · 2 comments · Fixed by #66
Closed
Assignees
Labels

Comments

@tomasjura
Copy link

When comparing collection (Enumerable) of objects which are not Strings, then result is incorrect. See attached example.

lcs_test1.rb.txt

@halostatue halostatue self-assigned this Jan 18, 2017
@halostatue halostatue added the Bug label Jan 18, 2017
@jagdeepsingh
Copy link

Pasting code and return values from link mentioned in the description:

require 'diff/lcs'

class M
  attr_reader :m

  def initialize(m)
    @m=m
  end

  def ==( o )
    o.m == @m
  end

  def hash
    @m.hash
  end

  def to_s
    return @m.to_s
  end    
end

s1 = %w(ma be c   e h j   l m      n po)
  => ["ma", "be", "c", "e", "h", "j", "l", "m", "n", "po"]

s2 = %w(ma be c d e f j k l m r s ta)
  => ["ma", "be", "c", "d", "e", "f", "j", "k", "l", "m", "r", "s", "ta"]

Diff::LCS::Internals.lcs(s1.map{ |s| M.new(s) } , s2.map { |s| M.new(s) })
  => [0, 1, 2]

Diff::LCS::Internals.lcs(s1, s2)
  => [0, 1, 2, 4, nil, 6, 8, 9]

@timon
Copy link

timon commented Jan 5, 2019

Struck me as well.

This is because the hash is used to map the objects to their positions.

mapping the strings to M objects ends up by instantiating different objects for the same string, and different slots are allocated in index hash for these objects.

The custom class should redefine eql? method to be an alias to == (in addition to custom == and hash methods) so that equal objects will use the same slot in hash.

halostatue added a commit that referenced this issue Jul 1, 2020
@halostatue halostatue linked a pull request Jul 1, 2020 that will close this issue
halostatue added a commit that referenced this issue Jul 1, 2020
# This is the 1st commit message:

Fix improperly placed chunks

Resolve #65

- Also add even more tests for checking `ldiff` results against `diff` results.
- Fix issues with diff/ldiff output highlighted by the above tests.
- Add a parameter to indicate that the hunk being processed is the _last_ hunk;
  this results in correct counting of the hunk size.
- The misplaced chunks were happening because of an improper `.abs` on
  `#diff_size`, when the `.abs` needed to be on the finding of the maximum diff
  size.

# This is the commit message #2:

Ooops. Debugger

# This is the commit message #3:

Restore missing test

- Fix some more format issues raised by the missing test.
- Start fixing Rubocop formatting.

# This is the commit message #4:

Last RuboCop fixes

# This is the commit message #5:

Finalize diff-lcs 1.4

# This is the commit message #6:

Fix #44

The problem here was the precedence of `or` vs `||`. Switching to `||` resulted
in the expected behaviour.

# This is the commit message #7:

Resolve #43

# This is the commit message #8:

Typo

# This is the commit message #9:

Resolve #35 with a comment
halostatue added a commit that referenced this issue Jul 1, 2020
- Resolve #65: Two different issues were reported:

  - Line numbers changed between 1.3 and 1.4. This change is intentional, but
    in comparing the output against `diff` for the new examples provided
    it became clear that for unified and context diffs, the last hunk was
    reporting incorrect hunk lengths (that is, if it was `12,4 11,8`, it should
    have been `12,3 11,7`. This has been resolved by providing
    `Diff::LCS::Hunk#diff` an additional parameter, `last` which defaults to
    `false`.

  - Net new lines were added in the middle of a unified diff hunk, which was
    incorrect. That is, we were getting:

    ```diff
    @@ -10,6 +11,10 @@
    recipe[j::default]
    recipe[k::default]
    recipe[l::default]
    +recipe[o::new]
    +recipe[p::new]
    +recipe[q::new]
    +recipe[r::new]
    recipe[m::default]
    recipe[n::default]
    ```

    instead of:

    ```diff
    @@ -12,3 +11,7 @@
    recipe[l::default]
    recipe[m::default]
    recipe[n::default]
    +recipe[o::new]
    +recipe[p::new]
    +recipe[q::new]
    +recipe[r::new]
    ```

    This has been resolved. The error was that `Diff::LCS::Block#diff_size` was
    applying `.abs` to its output, which is incorrect. We need to know the
    direction of the size for placing changes, but when determining the maximum
    block size for use in `Diff::LCS::Hunk#diff` calculations, we need to know
    the `.abs` size.

  - New tests were added to prevent these changes from regressing in the
    future, both as issue tests and as additional `ldiff` tests. These tests
    highlighted more issues with diff-lcs output as compared to `diff`,
    specifically the handling and reporting of missing newlines at the end of
    files. All of the issues highlighted were resolved and the structure of
    `ldiff` tests was changed so that it is easier to add new comparison files
    at any time.

- Resolve #35: Indicate that when comparing against custom objects, `#eql?`
  must be implemented such that objects that _resolve_ to the same meaning are
  _treated_ as the same meaning. This is important because the basic LCS
  algorithm uses a hash for position matching.

- Resolve #43: Provide a more meaningful error from `Diff::LCS::Hunk.new` if
  the `piece` provided does not create useful `Diff::LCS::Block`. It's
  extremely unlikely, this error will be more useful than `NoMethodError` being
  thrown.

- Resolve #44: `ldiff` binary detection failed to work correctly. It had been
  `!old_text or !new_text`, but the precedence of `or` broke that. It is now
  `!old_text || !new_text`.

Also:

- Ran Rubocop again. Updated configuration definitions, fixed some code
  formatting.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging a pull request may close this issue.

4 participants