Marek Gagolewski

I just recalled that environment variables can be passed to `install.packages` via `configure.vars`, so maybe this one: ``` Rscript --vanilla -e 'install.packages("stringi", configure.vars="http_proxy=\"http://www-proxy3.XXXX.XXX.de:8080/\"")' ```

Nice, I don't think it's a documented solution even :)

also should work for logical vectors ``` stri_subindex("abcde", c(T,F,T,F,T) == "acd" ``` single vector -> vectorized w.r.t. number of characters a list of integer or logical vectors can also be...

This can be imitated with #30, arg. `length == 1`

``` > stri_join_list(stri_sub_all("spam", c(4, 3, 2, 3, 1), length=1)) [1] "mapas" > stri_join_list(stri_sub_all("spam", list(c(4, 3, 2, 3, 1), c(1, -1)), length=1)) [1] "mapas" "sm" ```

It seems that R's `strptime` default to midnight local timezone today: ``` > strptime("2023 32", "%Y %M") [1] "2023-11-08 00:32:00 AEDT" ``` I will soon make stringi's behaviour compatible with...

Might be a good one! My three Aussie cents: * always use UTF-8 * consider including null-bytes at the end of each substring (makes processing by other functions easier; `the\0black\0cat\0`...

What about doing this at the R, not just package level? Maybe we should ping Tomas Kalibera and ask what he thinks about it... I'm a big proponent of unity..

I agree. We also need to think about how `NA_character_`s would be represented. NAs in the index vector (like `c(0,3,NA,8,11)`)? But then it would slow down string extraction for "sparse"...

A zero-length string with an additional NA-marker makes sense to me too. This will not paralyse other algorithms.