non equi join for character columns

[franknarf1]

I noticed this error message:

library(data.table)
data.table(id = LETTERS[1:3])[.(id = "B"), on=.(id > id)]
# Error in bmerge(i, x, leftcols, rightcols, io, xo, roll, rollends, nomatch,  : 
#   Only '==' operator is supported for columns of type character.

I couldn't find any open issue filed for adding this feature (since #1452 is done already). Is it not planned?

I ran into it looking at an SO question:

# goal: use bmerge for this subset condition: letter != '' & date < as.IDate('2014-10-01')
library(data.table)
dt <- data.table(letter = rep(c('',letters[1:4]), each = 4),
  date = as.IDate(c('2014-09-29','2014-09-30','2014-10-01','2014-10-02')))

# attempt
dt[.(L = '', D = as.IDate('2014-10-01')), on=.(letter > L, date < D), verbose = TRUE]
# Non-equi join operators detected ... 
#   forder took ... 0.02 secs
#   Generating non-equi group ids ... done in 0 secs
#   Recomputing forder with non-equi ids ... done in 0 secs
#   Found 4 non-equi group(s) ...
# Starting bmerge ...Error in bmerge(i, x, leftcols, rightcols, io, xo, roll, rollends, nomatch,  : 
#   Only '==' operator is supported for columns of type character.

(I figured this would work for letter != '' so long as "" is the minimum character.)

Anyway, not a big deal for me, but figured I'd ask.

[arunsrinivasan]

@franknarf1 != could work for char types, but isn't implemented yet. When done, this'll be taken care of. I can't remember if there's an issue for != addition to non-equi joins yet. IF not, the title of this issue could be changed to reflect that. I don't think it makes sense for other ops on char types.

Perhaps the error message could state that != when added will be made available for char types as well.

[franknarf1]

Thanks @arunsrinivasan , that makes sense.

Feel free to change the title (support bmerge for char !=?) or close as/when appropriate.

[jeroenjanssens]

I would like to be able to use <= and > for char types. Would that be as easy to implement as != ?

[jangorecki]

Are you aware of portability issues related to OS encoding? I would suggest to decode char to factor and non equi join integers. You will get what you need in a portable and reliable way.

[jeroenjanssens]

Thanks @jangorecki. I'm comparing strings of the format [0-9]{4} [A-Z]{2} so the many possible combinations make it infeasible to convert them to a factor.

I can imagine strings can be quite challenging to maintain portability. Do you know how base R does this when comparing strings?

[adamaltmejd]

Would be very interested to see "!=" join for character. Any progress on this? Or suggestion for workarounds meanwhile?