Fix #21242: Add REPL flag to quit after evaluating init script

[noti0na1]

Adding a new REPL flag --repl-quit-after-init: evaluate the init script and skip the interactive mode.

Fix the remaining part of #21242

Test with flag at REPL startup:

> ./bin/replQ --repl-quit-after-init --repl-init-script 'println("Hello from init script!"); val i = 2 * 2' 
Hello from init script!
val i: Int = 4
>

[Gedochao]

Can we have a test case for this?
Just to verify the output, and that a sub-terminal doesn't get created.
Preferably with dotty.tools.repl.Main, rather than the legacy runner.

[noti0na1]

Can we have a test case for this?

You can test it locally, the process terminates immediately, but I don't know how to create a test to verify that.

[Gedochao]

You can test it locally, the process terminates immediately, but I don't know how to create a test to verify that.

I can see that it works the way it is now, but an automated anti-regression test would go a long way.
I believe it'd be sufficient to call the REPL main method with the --repl-init-script and --repl-eval args, and then verify the output on stdout.

[noti0na1]

Oh, I find this test we may be able to reuse: compiler/test/dotty/tools/scripting/BashExitCodeTests.scala

[Gedochao]

Oh, I find this test we may be able to reuse: compiler/test/dotty/tools/scripting/BashExitCodeTests.scala

Seems the test is failing 🤔

[dwijnand]

Why --repl-eval? I really don't intuit "quit after evaluating init script" from that... Picking --repl-eval-init-script-and-quit is verbose but clear. But I don't know what the best name is... --repl-skip-interactive, --repl-eval-only is still iffy to me, dunno... 👨🏼‍🎨

[dwijnand]

Other than the name, LGTM

[som-snytt]

I had the same reaction to naming, but I haven't followed the conversation, which mentions Ammonite as a model.

Regular scala 2 REPL:

➜  snips scala -e 'println("hello")' -I world.scala
world
hello
➜  snips

If I wanted to switch the order of evaluation, I'd be inclined to add -E.

Edit: quick musing: there's a set of such things I must always look up, such as how to turn off colors in sbt vs scalac. I kind of remember, but it's impractical to try out all the variants, color vs colors, no vs never, etc. I can try out -E or -e quickly, if I've forgotten.

In a problem space where I can't try it out (I could try out ls -R vs ls -r if I forgot), then I will never use it casually. Possibly, if I used it, I would only cut/paste, unless I finally had to go in search of doc.

Another example is -Wconf, where I often must -Wconf:help. (Warnings now issue extra help.)

[noti0na1]

• scala -e only evaluates the file and show the output;
• --repl-eval will evaluate as the input to repl, so the output and result will be shown line by line.

Maybe @Gedochao can point out more usages 🤔

[noti0na1]

Why --repl-eval? I really don't intuit "quit after evaluating init script" from that... Picking --repl-eval-init-script-and-quit is verbose but clear. But I don't know what the best name is... --repl-skip-interactive, --repl-eval-only is still iffy to me, dunno... 👨🏼‍🎨

Or --repl-quit-after-init?

[Gedochao]

• scala -e only evaluates the file and show the output;
• --repl-eval will evaluate as the input to repl, so the output and result will be shown line by line.

Maybe @Gedochao can point out more usages 🤔

Generally speaking, it allows to interact with the REPL without spawning a sub-terminal, which will make testing REPL integrations (on top of which Scala CLI builds) much easier. That's also been the primary motivation for this change.
Other than that, it can be used rather similarly to scala -e.

[Gedochao]

LGTM

[Gedochao]

@noti0na1 not merging yet, as to let you finalize the naming of the option.
I'm fine with both what's currently there and what was suggested.

[hamzaremmal]

We should not ship the repl command, this diminishes SIP-46. No new script should go in the dist/bin folder without careful thinking (and it shouldn't have a generic name as repl).

[Gedochao]

We should not ship the repl command, this diminishes SIP-46. No new script should go in the dist/bin folder without careful thinking (and it shouldn't have a generic name as repl).

Right, I didn't connect the dots that this script would get subsequently shipped alongside scala and scalac.
100% agreed that we shouldn't ship it.

We all agree that we shouldn't ship a new script

[noti0na1]

• Rename the flag to --repl-quit-after-init so the behaviour is more clear by its name;
• Removed the script from dist/bin.