dfm.corpus(x, verbose = TRUE) never completes the verbose output

[kbenoit]

Unlike the tokens() sequence, where we always end with tokens.tokens(), dfm() does not end with dfm.dfm(). When verbose = TRUE, we don't get the ending.

Should look like:

> dfm(data_dfm_lbgexample, remove = "A*", verbose = TRUE)
Creating a dfm from a dfm input...
   ... removed 1 feature
   ... lowercasing
   ... created a 6 x 36 sparse dfm
   ... complete. 
Elapsed time: 0.011 seconds.

but instead we don't see the final two/three lines:

> dfmat <- dfm(data_corpus_inaugural, remove = stopwords("en"), verbose = TRUE)
Creating a dfm from a corpus input...
   ... lowercasing
   ... found 58 documents, 9,399 features
   ... removed 136 features

[koheiw]

Do we still need elapsed time?

Elapsed time: 0.011 seconds.

I want to remove these lines

quanteda/R/dfm.R

Lines 142 to 143 in d360d82

	dfm_env <- new.env()
	dfm_env$START_TIME <- NULL

If users are interested in timing, they just use system.time().

[kbenoit]

Yes I very much like the elapsed time and closing message. It tells us something about the timings and sequence, good for diagnostics and for some users who want to know the precise sequences and/or timing without wrapping the function, it's useful.

Works nicely for tokens() too.

> tokens(data_corpus_inaugural, remove_punct = TRUE, 
+        remove_symbols = TRUE, remove_numbers = TRUE,
+        verbose = TRUE)
Creating a tokens object from a corpus input...
...starting tokenization
...preserving hyphens
...preserving social media tags (#, @)
...tokenizing 1 of 1 blocks
...segmenting tokens
...serializing tokens 10062 unique types
...removing separators, punctuation, symbols, numbers 
...total elapsed:  0.406 seconds.
Finished constructing tokens from 58 texts.

[koheiw]

Have you ever really used the elapsed time for diagnosis? I don't think so, because elapsed time for single execution doesn't mean anything. When we want to find slow functions, we use profvis::profvis().

I like this part

...starting tokenization
...preserving hyphens
...preserving social media tags (#, @)
...tokenizing 1 of 1 blocks
...segmenting tokens
...serializing tokens 10062 unique types
...removing separators, punctuation, symbols, numbers

but not this

...total elapsed:  0.406 seconds.