Make watcher more simpler and more self-contained #149
Conversation
And only call it once, rather than once per evaluate call.
| #' @return list containing four functions: `get_new`, `pause`, | ||
| #' `unpause`, `close`. | ||
| #' @keywords internal | ||
| watchout <- function(debug = FALSE) { | ||
| watchout <- function(handler = new_output_handler(), |
There was a problem hiding this comment.
This is technically a breaking change, but (unsurprisingly) I don't see any evidence that anyone has ever used it. It might be worth un-exporting in the next version to make it very clear that it's for internal use only.
| watchout <- function(handler = new_output_handler(), | ||
| debug = FALSE, | ||
| frame = parent.frame()) { | ||
| con <- file("", "w+b") |
There was a problem hiding this comment.
A file() connection takes a bit longer to set up than a text connection, but it is faster to append too (especially as the output grows long since appending to a character vector is O(n^2)). It's also simpler and faster to read from, and any set up time is now amortised over all top level expressions.
R/watcher.R
Outdated
| }, | ||
| get_con = function() con | ||
| ) | ||
| read_con <- function(con, buffer = 1024) { |
There was a problem hiding this comment.
I think you can read more bytes from the buffer, e.g. 32k seems like a good number that allows a single read operation often.
And only call it once, rather than once per evaluate call.