tty: Deno.setRaw(rid, mode) to turn on/off raw mode

[kevinkassimo]

Difference from the first option:

• Deno.setRaw(rid, mode) to turn on/off raw mode
• Added Deno.isatty(rid) that can test any resources, not just stdin/stdout/stderr (personally prefer deprecating current Deno.isTTY())

Deno.setRaw(0, true);
Deno.setRaw(0, true); // safe

const buf = new Uint8Array(6);
for (let i = 0; i < 6; i++) {
  const nread = await Deno.stdin.read(buf);
  if (nread === Deno.EOF) {
    break;
  } else {
    console.log(`READ ${nread} bytes:`, new TextDecoder().decode(buf.subarray(0, nread)));
  }
}

Deno.setRaw(0, false);
Deno.setRaw(0, false); // safe

[ry]

I like the minimality of isatty(rid). Maybe you can do something similar here, instead of defining this new class TTYInput you could just add a new function... something like Deno.setRaw(rid). What do you think?

Additionally can you mark in the JSDOC any new APIs as

/** UNSTABLE: newly added API
  *
  *  Check if a given resource is TTY
  */
export function isatty(rid: number): boolean;

[kevinkassimo]

Do you think the original returning a restoration function after calling Deno.setRaw(rid) works here? (In case I want to store the information for restoring without exposing? This would be a pure TS side change, and similar to what I did in the old PR, but without wrapping in a class)

[ry]

I don't think storing the restore state is so important. The user can do that.

[kevinkassimo]

I don't think they can with current proposed API (unless we expose a full representation of <termios> utilities) if we do not store for them? And there is the discrepancy between Unix and Windows here

Node does not provide the API for direct restoration either. I believe with tty.ReadStream::setRaw(false) libuv simply sets back to the saved original mode

[ry]

libuv simply sets back to the saved original mode

Yes we should do that too... I mean that the JS interface does not need to expose state restoration just boolean on/off is good enough.

[kevinkassimo]

The slight annoyance is that we will need to accompany StreamResource (including FsFile) with such optional data. I'll see where would be the best place for me to place this piece of info.

[kevinkassimo]

Added a basic test (unix only, I have no access to Windows). Added pty crate as a dev dependency.

permissions::tests::test_permissions_request_read and permissions::tests::test_permission_request_net seems to be badly flaky locally

[ry]

@kevinkassimo Looks good but I'm a little confused because there is still Deno.isTTY() which if I'm reading your patch correctly does not call the new Deno.isatty() .. which I would expect it to.

[kevinkassimo]

@ry I can probably update isTTY to use isatty. I was originally thinking about deprecating isTTY since the 2 has very similar names while isTTY only handles stdin/stdout/stderr

[ry]

LGTM

[kevinkassimo]

Is this PR considered ready? Have been resolving conflicts with master updates for quite a few times 😅

[ry]

LGTM - sorry for the delay! Thanks!