Boxplot performance issues #119

Closed KelseyChetnik closed 3 years ago

In GitLab by @krumsiek on Jul 10, 2020, 11:53

Generating boxplots in mt_plots_boxplot and outputting them in mt_reporting_html takes way longer than it should, e.g. for 50-100 boxplots. Issue probably occurs when splitting into multiple pages. https://gitlab.com/krumsieklab/mt/-/blob/master/MetaboTools/R/mt_plots_boxplot.R#L191

In GitLab by @krumsiek on Jul 26, 2020, 14:50

Solution:

Minor speed-up: Parametrize to choose between (1) beeswarm plot, (2) jitter plot, (3) no sample plotting.
Substantial speed-up: Get rid of paginate calls and create a single facet_wrap. Get rid of row parameter (only col parameter is left, i.e. how many plots per row)
mt_reporting_generateMD.R needs to be adapted for adaptive heights (ask Elisa/Jan)
While implementing this issue, we should consolidate boxplot and scatter plot codes. https://gitlab.com/krumsieklab/mt/-/issues/113

In GitLab by @krumsiek on Jul 28, 2020, 15:22

jitter is already a parameter, see lines 174-177 of mt_plots_boxplot.R

In GitLab by @krumsiek on Jul 28, 2020, 15:26

to facet wrap: gg + facet_wrap(~var, scales="free_y", ncol = 4)

In GitLab by @elibene on Jul 28, 2020, 20:06

I would still add the option of using geom_jitter instead of geom_beeswarm, because it's much faster.

In GitLab by @krumsiek on Jul 28, 2020, 20:07

Yes, agreed. We will have all three variants, jitter, beeswarm, or none.

In GitLab by @krumsiek on Oct 6, 2020, 15:40

krumsieklab / maplet