BUG(web/table): Empty strings make sort misbehave, not sort properly about nextclade HOT 2 CLOSED

@corneliusroemer It seems that these values are undefined, not empty strings.

Nextclade Web uses stable sort (order of equivalent entries is preserved) and empty (undefined) values are not sorted, because we don't know how (JS might have some weird behavior from when it was slapped together hastily in 1992, but the correct computer-sciency way to say is that empty values are not comparable to strings). I tried to model the behavior for the well-known tabular software, such as Excel.

Consider this:

• you have fresh results from the Lapis link (perhaps fewer sequences, because 1000 is a bit too much to wait)
• you can sort by "Clade" column and all "outgroup" items will go to the top or bottom
• note that numbers in "i" column for "outgroup" entries have consecutive values - stable sort ensures that the order of values that all have equal "outgroup" clade is preserved the way it was before sort
• then if you subsequently sort by "Unaliased", the empty cells will stay at the bottom (or top), because we don't sort undefined values

Stable sort tries to preserve the existing order as much as possible and this allows you to sort/group by multiple criteria, for example by "outgoup" or not "outgroup" and then by lineage. Or you can first sort by lineage and then sort by clade - each clade group will then have lineages sorted within the clade. You can also try more than 2 criteria.

The fact that all "outgroup" entries have no lineage helps in this particular case, but technically it is a random coincidence. So if you think that we need to sort empty values, then we need to specify how. For example, we may consider them empty strings and if we sort lexicographically, they will be considered "less than" any other non-empty string.

I can also try and write another algorithm, but you'd need to specify it a little better than "sort misbehaves".

from nextclade.

I implemented sorting of empty values as empty strings in
#1152

@corneliusroemer please give it a try and tell me if this is the behavior you expected initially.

from nextclade.