Differing dendrogram outputs from plot - --labeltext is honored differently by dendrogram and matrix outputs

[Glfrey]

Hello,

I'm currently having an issue where sourmash plot is outputting two different dendrograms. I've attached an example and the code I used below. I get the same issue with version 4.0.0.(using for some old work) and the latest version.

sourmash compare -k=21 --output k21_all_compare --csv k21_all_compare.csv /home/b56j439/sourmash/all/sigs/*sig

sourmash plot k21_all_compare --labeltext new_labels.txt --csv k21_all_plot.csv

I've confirmed the new_labels.txt contains the same labels in the same order as the original label file, I just wanted to remove the directory structure from the labels.

[mr-eyes]

Hi,
To make sure I understand the issue. In summary, the only difference here is the tree coloring in the clustermap and the dendrogram, right?

Update: I can see the problem now. Thanks.

[Glfrey]

Thanks @mr-eyes , let me know if you need anything from me.

[Glfrey]

Hi @mr-eyes,

I have identified the problem, part of which was my fault ( I was missing the -labels argument) and the other part was due to asymmetric handling of the -labels argument for the two plots. I've implemented a fix and opened a pull request.

[mr-eyes]

Thanks @Glfrey
I will take a deeper look in a few days and update the issue.

[ctb]

ok, got it ;).

The problem is that the matrix plot is not observing the labels passed in by --labeltext without --labels, while the dendrogram plot is observing those labels.

One fix (that would not be backwards compatible necessarily) is to force args.labels to True if --labeltext is provided.

Related, it seems to me that having sourmash_fig.plot_composite_matrix take both show_labels and show_indices is confusing...

[ctb]

To test this all out, examine the dendrogram and matrix output of the following commands:

1. sourmash plot compare-demo - labels on dendrogram, indices on matrix (BUG)

2. sourmash plot compare-demo --labels - labels on both dendrogram and matrix ✅

3. sourmash plot compare-demo --indices - labels on dendrogram, no indices on matrix (BUG)

4. sourmash plot compare-demo --labels --indices - labels on both ✅

#2790 fixes this behavior.