23626 Text Compare auto-alignment still baffingly poor

[chrisjj]

E.g.


The far closer match with 343 is missed. Alignment is Patience Diff.

I've been reporting this on and off for years, and it never improves.

Surely this can't be working as designed.

[Aaron]

Each alignment algorithm will parse the file differently. If you switch to the other algorithms in the Session menu -> Session Settings, Alignment tab, do they work better for this data?

[chrisjj]

Each alignment algorithm will parse the file differently. If you switch to the other algorithms in the Session menu -> Session Settings, Alignment tab, do they work better for this data?

No. Should they?

[Aaron]

Hello,

Can you email sample files and include the BCSupport.zip from the Help menu -> Support; Export? I've tried creating a similar example file but the diff line is aligning to the other diff line. If emailing, please also include a link back to this forum thread.

[chrisjj]

Can you email sample files and include the BCSupport.zip from the Help menu -> Support; Export?

Done, for a redacted version:

I've tried creating a similar example file but the diff line is aligning to the other diff line.

Great. Then I look forward to hearing what I'm doing wrong. Thanks.

[chrisjj]

Aaron wrote in mail:

Enable the Folder Compare's Session Setting -> Alignment tab -> Use closeness matching (on by default), and that updates the comparison and aligns these similar lines.

Thanks. That works.



So what's the catch? I assume there must be one, else this would be on by default. Help doesn't say, or say how this works:

Mark the Use closeness matching checkbox to attempt to align the remaining mismatched lines based on their similarity.

[Aaron]

Use Closeness Matching is on by default, but could be disabled. The selectable methods are using specific algorithms, and Use Closeness matching is an extra bit of logic that we can add on top of that which we find usually helps but can be disabled by users who want a standard algorithm.

[chrisjj]

Use Closeness Matching is on by default

Thanks for the correction.

The selectable methods are using specific algorithms, and Use Closeness matching is an extra bit of logic that we can add on top of that which we find usually helps but can be disabled by users who want a standard algorithm.

So... what's the catch?

[Aaron]

It gives slightly different results, just like each of the algorithms handles different types of alignment. In most cases, it gives results closer to what a user expects, since it tries to find similar lines. Sometimes a user does not want to weight a similar section of line and, if so, they can disable it.

[chrisjj]

Thanks.