(When to) substitute labels for ids? (character matrices) about rnexml HOT 4 CLOSED

All your assumptions are correct. For example, when I convert a nexus file to nexml, the states (0/1) in the matrix end up as the values of the symbol attribute. For labels, you have to watch out a little bit to ensure that i) they are optional and not guaranteed to be unique (so don't use them as some sort of key or ID) ii) the label of the row or node should take precedence over the label of the otu, if both are available. You can imagine, for example, that the rows in an alignment have a label such as Genus_species_genbank_accession, while the otu is just Genus_species. We would like to retain both, in their relevant locations.

from rnexml.

@rvosa Great, thanks. I added labels the char elements in your example comp_analysis.xml file, after which we can do:

f <- system.file("examples", "comp_analysis.xml", package="RNeXML")
nex <- read.nexml(f)
get_characters(nex)

and get back:

         log snout-vent length reef-dwelling
taxon_8             -3.2777799             0
taxon_9              2.0959433             1
taxon_10             3.1373971             0
taxon_1              4.7532824             1
taxon_2             -2.7624146             0
taxon_3              2.1049413             0
taxon_4             -4.9504770             0
taxon_5              1.2714718             1
taxon_6              6.2593966             1
taxon_7              0.9099634             1

Note that

• the state symbols (0, 1) have been substituted for the state ids
• the otu labels taxon_n have been substituted for the otu numbers on the row elements, (as rownames)
• the char ids (colnames) have been substituted for the char labels.

As you point out, it sounds like we should be a bit careful when substituting labels. Currently, if no label is available, I keep the id number (otu id or char id) instead. Perhaps that behaviour is undesirable though, since it means the function is sometimes returning the char id and sometimes returning a label?

• Should we be using the label on the row itself (which I guess might read species_genus_accession) instead of the label on the corresponding otu for the rownames above?
• Or should we always return both the row label and the corresponding otu label as separate columns (even when one or both labels are absent or they are the same)? On one hand, this behavior seems the most consistent, but on the other, it's not likely to be convenient to most users (after all, we already have the extreme end of consistent but not convenient representation of the data as the S4 object).

@hlapp @balhoff Is my use of char node's label attribute consistent with the phenoscape expectation of where to find a 'human readable' column name? One might imagine that the columns should instead be labeled with abbreviation codes (e.g. LSVL instead of log snout-vent length, since spaces and dashes can be somewhat troublesome in column names), and then have the abbreviation defined in full elsewhere? This is the convention in the Ecological Metadata Language (EML); that columns are labeled by a short "attributeName" and defined by a longer "attributeDefinition" text string.

from rnexml.

@cboettig yes we use the label attribute for a human readable column name.

from rnexml.

On Nov 22, 2013, at 12:27 PM, Carl Boettiger wrote:

@hlapp @balhoff Is my use of char node's label attribute consistent with the phenoscape expectation of where to find a 'human readable' column name? One might imagine that the columns should instead be labeled with abbreviation codes (e.g. LSVLinstead of log snout-vent length, since spaces and dashes can be somewhat troublesome in column names), and then have the abbreviation defined in full elsewhere? This is the convention in the Ecological Metadata Language (EML); that columns are labeled by a short "attributeName" and defined by a longer "attributeDefinition" text string.

I don't think there's a convention stating that the char node label needs to be a short string or conversely can't be one. In Phenoscape this is just taken from the character matrix. Systematists don't ever call a character "LSVL" as far as I've seen. But that doesn't mean that other sources for matrices can't be using a short name that's expanded elsewhere.

If there is only one form, I'd suggest using rdfs:label. If there are two, a short form serving as column label and a long form giving a human-readable definition, I'd suggest to use refs:label for whatever is supposed to be the column label, and dc:description for whatever its explanation is.

from rnexml.