dfm_compress() error

[amatsuo]

For a large dfm, dfm_compress() generates an error. Here is a replication code (kbenoit has the data).

mentions_tab <- dfm_select(data_dfm_brexittweets_junesample, "@*") 
colnames(mentions_tab) <- sample(letters, size = ncol(mentions_tab), replace = T)
    #sub("^(@.+?)\\W.+", "\\1", featnames(mentions_tab))
mentions_tab <- dfm_compress(mentions_tab, margin = "features")

[kbenoit]

It's an error caused when a document has no features:

testdfm <- new("dfmSparse", Matrix::Matrix(matrix(c(0,0,0, 2,1,0, 0,1,0, 1,1,0), nrow = 3,
                                       dimnames = list(docs = paste0("d", 1:3),
                                                       features = c("a", "b", "c", "a"))),
                                sparse = TRUE))
testdfm
## Document-feature matrix of: 3 documents, 4 features (58.3% sparse).
## 3 x 4 sparse Matrix of class "dfmSparse"
##     features
## docs a b c a
##   d1 0 2 0 1
##   d2 0 1 1 1
##   d3 0 0 0 0

dfm_compress(testdfm)
## Error in validObject(r) : 
##  invalid class “dgTMatrix” object: length(Dimnames[1]) differs from Dim[1] which is 2