Kenneth Benoit

What if we use that code to add the following argument to `tokens_select()` and `dfm_select()`: ```r tokens_select(x, match_components = TRUE) dfm_select(x, match_components = TRUE) ``` which matches `pattern` to each...

The problem is not that we cannot solve this through code like that above, but rather that this resembles a StackOverflow answer, not the sort of convenient and robust solution...

@amatsuo Can you investigate this? Let's discuss how this will happen prior to making any actual changes.

Yes [bfg](https://rtyley.github.io/bfg-repo-cleaner/) works well for this (`brew install bfg` is the easy way to install on macOS), but might involve pushing to a new repo, which we would rename to...

Probably better to avoid non-CRAN packages for now, but we can definitely remove them from the git history and then .gitignore them so that they are never added back.

I'm hesitant to burn the past entirely, but would be ok identifying a relatively modern release and truncating from that. What will this do to previous issues/releases/tags/etc?

Thanks for working on this, at wit's expense even. We can close the PRs, since the branches will still be there, and we can reopen them up later, if this...

OK, so it sounds like we have implemented this in #1141. This converts a **quanteda** `tokens` object into what you have called here `intermediate_results`. We cannot implement the options in...

See the discussion in PR #1141, which we will reopen when we return to this issue.

This follows on https://github.com/quanteda/quanteda/pull/1141#issuecomment-457817537, but I prefer the discussion to take place here, on an open issue rather than a closed pull request. How about these: #### for a dfm...