Proof of concept for sharing Babel transpilation across processes

$ git clone [email protected]:jamestalmage/__share_transpile_across_process.git
$ cd __share_transpile_across_process
$ node test1.js

# Output

....................................................................................................
Launched server, loaded babel, plus:
100 children forked and loaded two transpiled dependencies each in 3324ms.

For comparison

$ node comparison.js

....................................................................................................
Loading up an entire babel instance per fork: 
100 children forked and loaded two transpiled dependencies each in 16459ms.

Using a long running background server on Dev machines improves the situation more

# launch the background server
$ node server

# switch to a new terminal tab

$ node test2.js

# Output
            
....................................................................................................
Using background server:
100 children forked and loaded two transpiled dependencies each in 2918ms.

Further Enhancements: Forking child processes is costly. Advanced coordination might improve efficiency further:

• Static code analysis for require calls, and batching with known dependencies in the same response.
• For large test suites, cache to disk and provide forks with a list.
• Caching between test runs (using last modified metadata on files).

Debugging

If something is not working launch the server (node server.js) and then launch a single run of forked-child

$ node forked-child.js $(which curl)

License

MIT © James Talmage