Memory #708
Memory #708
Conversation
isolate renderLeaflet so the map isn't redrawn when the region$plot changes
Now region$plot is reactive these aren't needed
This means we store the processed files globally
|
I navigated to Hampshire on multiple different browser windows and recorded memory usage in R by
So now you can see multiple sessions adds hardly anything to the total MB of R objects! I've added a cap so that objects are removed from the global list if there are more than 100 and they are removed last accessed == first removed. This is because I don't want that global list to get too large (loading all the rds files is ~ 15Gb). This will need testing as I've also made |
|
Thanks @nikolai-b I just took a look at the code, looks clever. Look forward to testing this when back home next week. Have asked a few questions in the meantime. I suggest that if it passes informal tests by me and @usr110 we get others to test it on the test server with a before/after merge. Sound reasonable? I think this could make the app way more tolerant of many users so great work! |
|
I've tested on the latest version with the new tiles in all scenarios for commuting and schools and for multiple regions. Aside for making large regions load faster, I think, I can see no change. So merging. Great work @nikolai-b, many thanks! |
| filepath <- file.path(base_path, filename) | ||
|
|
||
| while (length(loaded_data) > 100) { # Rm objects from the list when too many (by time last accessed) | ||
| loaded_data[[names(loaded_data_accessed[loaded_data_accessed == min(unlist(loaded_data_accessed))])]] <<- NULL |
There was a problem hiding this comment.
What does min(unlist(loaded_data_accessed)) here do?
| rds <- readRDS(filepath) | ||
| if(filename == "rnet.Rds") { | ||
| if(!is.null(rds)){ | ||
| rds$id <- rds$local_id |
There was a problem hiding this comment.
Sorry you've had to do that. We should have called it id initially I think but not worth changing now.
| if(filename == "rq.Rds"){ | ||
| # Merge in scenario data for quiet routes - don't want this in download but need for line sorting | ||
| rds@data <- cbind( | ||
| rds@data[!(names(rds) %in% names(str_lines))], |
There was a problem hiding this comment.
Shows that shifting from sp to sf would be a massive operation. Not worth it now.
Ref #707
This isn't ready but it shows the basic idea.