textmodel_wordfish raises error for large dfm

[cschwem2er]

Hi,

I have a dfm large 22.300x10.047 corpus, for which fitting a wordfish model raises the following error:

Error in wordfishcpp(as.matrix(data), as.integer(dir), 1/(priors^2), tol, : std::bad_alloc
3.
stop(structure(list(message = "std::bad_alloc", call = wordfishcpp(as.matrix(data), as.integer(dir), 1/(priors^2), tol, disp, dispersionFloor), cppstack = NULL), .Names = c("message", "call", "cppstack" ), class = c("std::bad_alloc", "C++Error", "error", "condition" ...
2.
wordfishcpp(as.matrix(data), as.integer(dir), 1/(priors^2), tol, disp, dispersionFloor)
1.
textmodel_wordfish(dfm14)

I checked the RAM of my PC during the function call and there were over 8GB left right before the error. Is this related to #124? Does wordfish still require a dense matrix?

[kbenoit]

I think this is the same as #124. @HaiyanLW is working on this fix.