hook into `tsc --watch` using stdout -- the unix way

[ORESoftware]

This is a feature request.

I would like to launch tsc -w and then read from stdout to know when tsc -w has started and completed a build. Like so:

const cp = require('child_process');

const k = cp.spawn('bash');

setImmediate(function(){
     k.stdin.end(`\n  tsc --watch  --project x  \n`);      
});

let stdout = '';
k.stdout.on('data', function(d){
    stdout += String(d);
    if(/foobar/.test(stdout)){
        makeMyDay();
    }
});

Right now, tsc --watch outputs stdout that is mostly just human readable, and doesn't provide very useful information:

1:33:44 PM - File change detected. Starting incremental compilation...

../suman-types/dts/it.d.ts(6,18): error TS2430: Interface 'ITestDataObj' incorrectly extends interface 'ITestOrHookBase'.
  Property 'cb' is optional in type 'ITestDataObj' but required in type 'ITestOrHookBase'.
../suman-types/dts/test-suite.d.ts(13,23): error TS2503: Cannot find namespace 'Chai'.
1:33:44 PM - Compilation complete. Watching for file changes.

Proposal:

tsc --watch should have a new flag, that tells tsc --watch to output machine readable data to stdout, instead of human readable data. For example:

tsc --watch --machine-stdio

Using this new flag, no existing users will be affected.

Ideally, output JSON to stdout, something like this upon a file change:

const data = {source: '@tscwatch', event: 'file_change', srcfilePath: filePath, willTranspile: true/false, destinationFilePath: null / destinationFilePath};
console.log(JSON.stringify(data);

then after transpilation finishes, write something like this to stdout:

const data = {source: '@tscwatch', event: 'compilation_complete', errors: null / errors:[]};
console.log(JSON.stringify(data));

if it's behind a flag (--machine-stdio) then it won't affect any current users, so should be safe.
I need a truly unique field value like '@tscwatch' so that I know that the JSON is coming from a certain process. In general, it's possible to be parsing more than one JSON stream from a process, so having a unique field in each JSON object makes this better.