How are we handling rank validation? about taxa HOT 4 OPEN

I like dealing with either lists or data.frame's of the options you gave.

Cool. I like the data.frame the most since it avoids numbers/NA as names.

I lean towards putting this off until next cran push

fine with me

from taxa.

Makes sense to do validation of ranks when database is specified and be specific to a database. ✔️

For order checking, if the ordering is wrong, do we error with message giving the correct order, or correct the order with message?

what is going on with the replication in ranks_ref

I believe i combined all the diff. rank names from diff. data providers, and just left duplicates in, they can def. be taken out.

Do you think this is important to get done before cran push?

from taxa.

For order checking, if the ordering is wrong, do we error with message giving the correct order, or correct the order with message?

I think correcting the order with a message makes more sense.

Thinking about this more, there might be some valid ranks that do no have an order (e.g. "unranked") and can therefore be put in any order relative to ordered ranks (e.g. "species").
This makes correcting for order and encoding it more difficult.
I dont think an ordered factor would work since AFAIK you cant make a ordered factor with some values ordered and other not. Below are the options for encoding the valid ranks in the database class I came up with that handle a mixture of ordered and unordered ranks.

vector

> c("1" = "root", "2" = "domain", "3" = "kingdom", "NA" = "unkranked")
          1           2           3          NA 
     "root"    "domain"   "kingdom" "unkranked" 

list

> list("1" = c("root", "unkranked"), "2" = c("domain", "unkranked"), "3" = c("kingdom", "unkranked"))
$`1`
[1] "root"      "unkranked"

$`2`
[1] "domain"    "unkranked"

$`3`
[1] "kingdom"   "unkranked"

data.frame

> data.frame(order = c(1:3, NA), rank = c("root", "domain", "kingdom", "unkranked"))
  order      rank
1     1      root
2     2    domain
3     3   kingdom
4    NA unkranked

Instead of the NA, we could replicate "unranked" for each level as was done for the list above.

two categories

ordered_ranks  <- c("root", "domain", "kingdom")
unordered_ranks <-  c("unkranked")

Do you like any of those in particular or have another solution?

, they can def. be taken out.

Cool

Do you think this is important to get done before cran push?

Its not essential. If it turns out to be an easy change then I think it would be nice, but leaving it out would not pose any backward compatibility issues since we can use database-less ranks for now.

from taxa.

I like dealing with either lists or data.frame's of the options you gave.

I lean towards putting this off until next cran push

from taxa.