use actual temp files and directories

[brentp]

currently we use /tmp/x.html. should use a real temp file and temporary directory.

[timotheecour]

yes but that's not enough; currently lots of things are broken by fact x.html is deleted right after browser opens it:

• curl file:///private/tmp/x.html doesn't work
• refresh doesn't work (gives 404)
• chrome => view page source doesn't work

proposal 1 (trivial)

• expose in nim API user-configurable file where html is saved (instead of result = "/tmp/x.html")
• if provided, use it and do not delete it after it was written to; else use current logic

proposal 2 (in addition to proposal 1)

when user didn't provide that file path:

• set it to `os.getTempDir()/"nimplotly"/plots_<something_random>
• (maybe) use https://github.com/OpenSystemsLab/tempfile.nim for the temp files

proposal 3 (if needed, down the line...) tempfile garbage collection

note: this is a generally useful feature that should be implemented in another package

• Use some temp file garbage collection to periodically remove the oldest temp files that were created
• an index.json is created/updated in same dir as temp files being created is updated each time a temp file is added
• the index.json specifies N=the max number of files kept; when that number is exceeded, the oldest temp files are removed (oldest in the sense of lastAccessTime for OS that support it, maybe), so that no more than N files remain
• the index.json can also have a reference to files that shouldn't be removed
• a nice nim API abstracts all of this into something nice and simple like:

let tempFile = createManagedTempFile(config = Config(dir = getTempDir(), maxFiles = 100, ...))
``

[Vindaar]

I think proposal 2 for a start would be great. I agree that proposal 3 would be nice, but as you said, it should be a separate library.

I'd argue that for most use cases not removing files shouldn't be an issue. The space needed even for > 100 plots is still small. In use cases where this is an issue, I would think the user would not go with a JS based plotting library anyways.