Use spawn

[renkun-ken]

What problem did you solve?

This PR makes use of spawn and spawnSync across the entire project. In particular, getAliases is updated to use spawnSync to be consistent with others.

[ManuelHentschel]

Thanks for the work. If I remember correctly, we only use a file to receive the results of getAliases.R, since there were some buffer overflow issues with exec. These should not be an issue with spawn according to e.g. https://stackoverflow.com/a/56688608, so I it might make sense to use stdout again. If you agree, I could implement that later today.

[renkun-ken]

Personally, I prefer receiving small data (a file path, a small chunk of text, etc.) directly from stdout whereas large data (all aliases, all rmd template info, etc.) from a temp file. Here we use spawn and spawnSync for consistency across the project in terms of command argument quoting, shell wrapping, etc.

Does that make sense?

[ManuelHentschel]

If there are no issues with buffer overflows, are there advantages to using a file over stdout? We load the entire file into memory at some point, anyways, and using temp files always has some potential to cause problems with cleanup, access restrictions, or even performance (on very limited systems).

[renkun-ken]

It looks like spawnSync also has a buffer size limit yet it could be adjusted. From spawnSync documentation, the buffer size could be specified in options:

maxBuffer: Largest amount of data in bytes allowed on stdout or stderr. If exceeded, the child process is terminated and any output is truncated. See caveat at maxBuffer and Unicode. Default: 1024 * 1024.

Do you mean we better use spawn and collect the stdout until the process exits?

[ManuelHentschel]

It looks like spawnSync also has a buffer size limit yet it could be adjusted. From spawnSync documentation, the buffer size could be specified in options:

maxBuffer: Largest amount of data in bytes allowed on stdout or stderr. If exceeded, the child process is terminated and any output is truncated. See caveat at maxBuffer and Unicode. Default: 1024 * 1024.

Do you mean we better use spawn and collect the stdout until the process exits?

Yes, that's what I had in mind (assuming there is no buffer with the async version)

[renkun-ken]

Now I switch to using spawn.

[ManuelHentschel]

Is there a reason that we show some errors to the user and log the errors in the childprocess/non-matching regex only to the console?

[renkun-ken]

Now an error message is shown to user.

I'm wondering if it makes sense to use a global output channel to show details of such information? But we already have multiple output channels: R Language Server, R Markdown, and VSCode-R-Debugger also has R Debugger.