TODO.md

TODO

Random observations, things that could be fixed.

download_table_TPMS_without_means is strange

It uses the global variable 'tissues' and ignores the tissue selection in the UI. This is probably unintentional.

The code is also very convoluted and could probably be replaced with a single SQL statement.

zooming doesn't work

If a region is entered and "Update Transcripts" clicked, the diagram isn't updated to the region. Instead, transcripts overlapping the region are queried, and the displayed region is the smallest one that includes all of them.

If this was changed to display an arbitrary region, the diagramming code would need to cope with transcripts that don't fit entirely into the diagram.

only one strand is shown

By default, only transcripts on one strand are shown, and the display is flipped so transcripts run left-to-right. It might make sense to show antisense transcripts, too, but the diagram should clearly indicate theyr direction.

convoluted code that could be SQL

• calculate_inner_exons
• download_table_TPMS_without_means
• get_TPM_from_tissues
• get_TPM_from_tissues_over_transcripts

get_transcripts_from_gene needs refactoring

Depending on how it was triggered, this monolith of a function does six similar, but subtly different things. Due to copy-and-paste coding, some of the differences are probably unintentional. They are also unexpected from a user perspective.

Commonalities should be factored out, redundancies removed, and similarity be enforced through code reuse.

get_group_comparisons_over_transcripts looks fishy

The code doesn't seem to know whether groupA and groupB are supposed to be lists or strings containing an encoded list. It first treats it as a list, then tries to split the string into a list. This can't be right, and the splitting code is awful, too.

duplicated and unclean logic

It is possible to select samples or tissues for grouping, or even a mix of the two. Multiple code paths then switch on whether the first selected item is a sample and then run queries that match on either samples.tissue or samples.name. Mixed groups are never handled correctly.

A clean way out could be to always treat the groups as mixed and match samples on either name or tissue. That should shorten the code and make it easier to keep consistent.