Explicitly Clustering One Axis When the Other is Already Clustered Ends Up Dropping the Other's Tree

Issue #355 closed
Robert Leach created an issue

USE CASE: WHAT DO YOU WANT TO DO?

Keep the existing clustering of (e.g.) columns and apply new clustering to (e.g.) rows.

STEPS TO REPRODUCE AN ISSUE (OR TRIGGER A NEW FEATURE)

  1. Open an unclustered data file
  2. Cluster only the columns by selecting "Do not cluster" for the rows in the cluster dialog
  3. Perform clustering a second time, but this time, select "Do not cluster" for the columns in the cluster dialog

CURRENT BEHAVIOR

After step 3, you end up with a matrix that appears clustered on both axes (which is good/expected), but the column tree goes away.

EXPECTED BEHAVIOR

If the user selects "Do not cluster" for an axis that has already been clustered and shows a tree, that tree should be retained after clustering the opposing axis.

DEVELOPERS ONLY SECTION

SUGGESTED CHANGE (Pseudocode optional)

If a user selects "do not cluster" when a tree exists for that axis, copy that tree into the new output files structure.

FILES AFFECTED (where the changes will be implemented) - developers only

unknown

LEVEL OF EFFORT - developers only

minor

COMMENTS

This issue is similar, but I suspect significantly different from issue #256.

I am unsure, but it might be the case that issue #256 solves this issue as a side-effect. The description of the issue in #256 however does not make this specific case.

This image shows the described result. Unfortunately, I did not take a screen cap after step 2, but it would have shown a column tree and no row tree as expected.

cluster_rows_only_when_cols_already_clustered-col_tree_goes_away.png

Comments (6)

  1. Christopher Keil repo owner

    This issue has definitely been fixed as of #256 and it was very much related to the issue that caused the original problem of #256. The updated master branch does not have this issue or #356 anymore. I am unable to reproduce it with both Windows and Linux and testing multiple files and I am 100% sure that this issue was part of resolving #256.

  2. Robert Leach reporter

    I just tried to confirm this was fixed in the current master and hit an exception. I created issue #362 to address it. Once that's fixed, I will test this again to see if I can mark it as resolved.

  3. Log in to comment