feat(vis-type: scatter): improve rendering of scatter plot

[dvmoritzschoefl]

Closes #522 ,https://github.com/datavisyn/reprovisyn/issues/1870, https://github.com/datavisyn/communication_bi_ordino/issues/76, https://github.com/datavisyn/communication_bi_ordino/issues/77,
https://github.com/datavisyn/reprovisyn/issues/2028,
https://github.com/datavisyn/communication_bi_ordino/issues/79

Developer Checklist (Definition of Done)

Issue

• All acceptance criteria from the issue are met
• Tested in latest Chrome/Firefox

UI/UX/Vis

• Requires UI/UX/Vis review
  • Reviewer(s) are notified (tag assignees)
  • Review has occurred (link to notes)
  • Feedback is included in this PR
  • Reviewer(s) approve of concept and design

Code

• Branch is up-to-date with the branch to be merged with, i.e., develop
• Code is cleaned up and formatted
• Unit tests are written (frontend/backend if applicable)
• Integration tests are written (if applicable)

PR

• Descriptive title for this pull request is provided (will be used for release notes later)
• Reviewer and assignees are defined
• Add type label (e.g., bug, feature) to this pull request
• Add release label (e.g., release: minor) to this PR following semver
• The PR is connected to the corresponding issue (via Closes #...)
• Summary of changes is written

Summary of changes

• fix regression stats

Screenshots

Additional notes for the reviewer(s)

Thanks for creating this pull request 🤗

[thinkh]

@dvmoritzschoefl as discussed in call:

1. Distribute subplots equally for the available width
   

2. Changing the drag mode in the vis config is not reflected in the plot

[dvmartinweigl]

Could we maybe expose a prop in the scattervis config to change the type of splom.
What do you think @dvmoritzschoefl

Something like this:
splomMode:

• lowerTriangle
• fullMatrix
• hideDiagonal
• histogram (if added in future)

[dvmoritzschoefl]

Could we maybe expose a prop in the scattervis config to change the type of splom. What do you think @dvmoritzschoefl

Something like this: splomMode:

• lowerTriangle
• fullMatrix
• hideDiagonal
• histogram (if added in future)

I would not make it more complicated than needed for now. I think if we cannot have histograms, then the lower triangle mode is the best bet (and also the best performing one). I would add it if someone needs this at some point.

[dvmoritzschoefl]

@dvmoritzschoefl as discussed in call:

1. Distribute subplots equally for the available width
   
2. Changing the drag mode in the vis config is not reflected in the plot

The number of columns and rows is now calculated by the following rule:

• per 400 pixels one more columns comes (until a maximum of 4 columns) -> if less facets than columns are present then the number of columns is the same as the number of facets
• the number of rows is facetnumber / columnnumber
  while not perfect for sure this works well for the 1-10 facet cases

as for the plot not updating and the dragmode having no effect. i think this is an ordino problem (as discussed in call). i tried it in the storybook and when using the config in an external state and updating there it works as expected

[dvchristianbors]

One more small remark is regarding the legend and coloring of unknown categorical values

The categories are not 'sorted' (previously, the 'Unknown' category always came last), and the vis_neutral_color is not used, as recommended:

/**
 * Neutral color (e.g., histogram in scatterplot matrix and should be also used for "Unknown" categorical values)
 */
export const VIS_NEUTRAL_COLOR = '#71787E';

https://github.com/datavisyn/visyn_core/blob/a2d4450bd2eb9a5d86d84fb4ac99fe0ba1c5c15c/src/vis/general/constants.ts

#71787E

[dvmartinweigl]

Something seems wrong with the stats calculations. Adding 4 numeric columns (all gene expression) to the config results in the last row of the lower triangle having NaN values for pValue, r² ....
Could it be that there is an error in the data getting passed to fitRegressionLine @dvmoritzschoefl ?

[thinkh]

Thanks for the performance improvements and addressing all the various feedback. We have tested it in multiple apps and it is ready to merge.

[thinkh]

@dvmoritzschoefl I saw that the Playwright tests are failing for this branch. in comparison the develop branch succeeded with the latest merge. Please check the tests and fix them. Thanks.

[thinkh]

I fixed the failing Playwright tests by adding and using the correct selectors for the legend.

[thinkh]

The download Playwright tests are currently timing-out but all other tests are working as expected.