Add select_ to support later dplyr

[jeffbaumes]

There are two problems when supporting R version 3.1:

• The string of conversion seems to act differently, so I replaced it with no conversion (seems to work when data is a dataframe) dat <- tbl_df(as.data.frame(lapply(1:ncol(data), function(x) type.convert(as.character(data[,x])))))
• dplyr now contains and uses a select_ function, and I put one in that appears to work ok (though there could still be issues with it)

Still to resolve

• How to do the conversion correctly for R 3.1
• How to export select_.treedata for R 3.0 when select_ is not defined. I attempted this in an else block, but someone would need to confirm on R 3.0 since I upgraded my system and now don't really want to go back :) - Or maybe we just don't support R 3.0 anymore and force people to upgrade.

[jeffbaumes]

@uyedaj @lukejharmon Please take a look. I may be a bit out of my league in R programming but I gave it a shot.

[jeffbaumes]

I honestly have no idea what that conversion line does, so I didn't decompose to find the true issue. Indeed rownames seem to disappear somewhere in the conversion. I have also noticed that both in select_ and near the end of the make treedata there is a line that attempts to set rownames, but there is no effect (row names remain unset) and I have no idea why. If those lines worked, the conversion line could be kept in and the row names would just be set later. But as it is, on R 3.1 removing that conversion was the only place I could figure out how to keep rownames in place for downstream processing.

[uyedaj]

Ok, that makes sense. I'll look into how to make this more stable...