Add str_select which works as grep with value=T

[jiho]

The code is actually based on str_detect at the moment. It could wrap grep if
this is preferred.

[jiho]

Any chance of getting this in? I am using x[which(str_detect(x, ...))] more than I can count.

[matthieugomez]

Do you really need which?

string[str_detect(string,pattern)]  

[jiho]

Yes because:

str_detect(c(NA, "abc"), "a")

gives NA and which eliminates those.

[matthieugomez]

Thanks!

[hadley]

@gagolews do you think this would be useful for stringi?

@jiho I'd prefer to use is.na(x) & x, but mostly sorry for taking so long to look at this.

[gagolews]

@hadley: Do you mean stri_subset_* here? Its already there (?)

[hadley]

Oops, missed that. I'll add a wrapper!

[hadley]

Done!

[jiho]

Thanks for the regain of activity on stringr! I did not know about stringi, it looks cool. The packages seem very similar now. Notwithstanding company policies (RStudio vs. Rexamine), wouldn't it save you both a lot of time to just merge them? Having several packages that do almost the same thing is also an aspect of R which is difficult for new / unconfident users. Anyhow, thanks for the work.

[hadley]

@jiho two main reasons:

• Updating stringr gives every existing stringr user a big performance boost
• stringi is more of a developer library - I think it provides so many options that it would be overwhelming to a new user. So stringr provides a simple high-level API that glosses over some of the details in the interest of simplify, where stringi dives deep, giving you the ability to do anything you could ever imagine.

[jiho]

@hadley thanks for the info. I'll surely continue to use stringr since I am comfortable with it and most of my code uses it. I was just wondering wether it was OK for stringi to be "relegated" to work behind the scene in that configuration, the ICU core might as well be integrated into stringr directly, adding some new functions for the complex aspects.

But anyhow, as a stringr-only user, I'm happy!