ropensci-archive/binomen

:no_entry: ARCHIVED :no_entry: 'Taxonomic' Specification and Parsing Methods

binomen's Issues

to jsonld (xml too?)

Fxn started, but not working yet:

• try to follow darwin core terms
• meh, for XML, we'd need XML pkg, maybe just stick with jsonld

Fill out various functions to work on taxa class objects

In addition to on taxondf and taxon objects

Thoughts on binomen

Hi @sckott, you asked for my thoughts on binomen on issue #9 and I thought a new issue might be a better place to respond.

I like the package overall. It has a clean and straight forward style that I think biologists will like.
Below is a list of intermixed comments, suggestions, and potential bugs.

• The classes assume explicit ranks. I have mixed feelings about this. On one hand, this is intuitive for biologists who might want to characterize data by a familiar rank, like family or genus. Having defined ranks means that you know the order (e.g. does 'aberration' come before 'subform'?). This also gives you a way to force all kinds of data into a table or equal-depth taxonomy, however incomplete, since you know the dimension corresponding to rank. On the other hand, data from some sources, such as database FASTA headers, might have the taxon names without their ranks and some people might be fine not thinking about the ranks; forcing those people to look up the ranks via database searches might be too much of a hassle for some. Or, ranks might be available, but the rank names/order might be different than what is defined in the package, forcing the user to figure out how to convert their raw data before forming it into binomen classes. Also, genus must be defined, yet in some cases you might not know the genus, such as ambiguously classified OTUs. Overall, it seems to me explicit ranks are more intuitive, but less flexible. I lean towards arbitrary/optional ranks.
• When there is no species epithet, none is printed. How about printing this as sp.?
• When no species is defined none is printed twice in the species field of the binomial. Is this supposed to happen?

> obj <- make_taxon(genus="Poa", family='Poaceae')
> obj$binomial
<binomial>
  genus: Poa
  epithet: none
  canonical: Poa none
  species: Poa none none
  authority: none

• For span it would be nice to be able to leave off one end of the range (e.g. df2 %>% span(family, NA) or df2 %>% span(family, )), meaning that all ranks above/below the specified rank are included.
• You have to specify genus to make a taxon, but not when making a taxondf. This can cause an error when trying to convert taxondf to taxa:

> df <- data.frame(order = c('Asterales','Asterales','Fagales','Poales','Poales','Poales'),
+                  family = c('Asteraceae',NA,'Fagaceae','Poaceae','Poaceae','Poaceae'),
+                  genus = c('Helianthus', NA,'Quercus','Poa','Festuca','Holodiscus'),
+                  stringsAsFactors = FALSE)
> df2 <- taxon_df(df)
> scatter(df2)
Error in if (genus == "none") stop("You must supply at least genus") : 
  missing value where TRUE/FALSE needed

Allow filter with operators < >, etc.

We have span, e.g., df2 %>% span(order, genus) and takes two rank names to get range of data from

Would be nice to allow searches like

• obj %>% select( > family)
• obj %>% select( < genus)
• obj %>% select( != tribe)
• obj %>% select( == tribe)

though this would already work with dplyr for data.frame's but not on taxon/etc. objects of course...hmmmm

Maybe include proper Darwin core names as part of taxon class

http://rs.tdwg.org/dwc/terms/index.htm#Taxon

clazz?

Minor question. There are multiple occurrences of clazz in the README.md. Should this be "class", or are you trying to avoid using the reserved word "class"?

basic vignette

code of conduct

Prob. rename classification() to avoid conflict with taxize::classification