Skip to content
This repository has been archived by the owner on Sep 16, 2021. It is now read-only.

Get JSCompiler to compile some TS code #74

Closed
wants to merge 3 commits into from

Conversation

achew22
Copy link
Member

@achew22 achew22 commented Nov 19, 2017

Months of teath nashing later...

Very far from perfect, but it produces a bundled JS. Launch and iterate... right?

Fixes #6

@achew22 achew22 force-pushed the closure_compiler branch 2 times, most recently from 26e6e62 to d900a87 Compare November 19, 2017 22:23
# Bazel < 0.8.0.
"bazel-out/local-fastbuild/bin/",
# Bazel >= 0.8.0.
"bazel-out/k8-fastbuild/bin/",

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

On OSX it seems to be darwin_x86_64-fastbuild. Maybe you could add this.

Copy link
Contributor

@mrmeku mrmeku left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just some comments of how to address a todo and how to clean up the code a bit

"--dependency_mode=LOOSE",
"--js_module_root=bazel-out/k8-fastbuild/bin/examples/closure_compiled/examples/closure_compiled",
"--jscomp_off=reportUnknownTypes",
"--jscomp_off=missingSourcesWarnings",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This should be the only def you need to keep. Remove the rest

@@ -301,6 +327,49 @@ def compile_ts(ctx,
collect_default=True,
collect_data=True,
),
"closure_js_library": struct(
# File pointing to a ClosureJsLibrary protobuf file in pbtxt format
# that's generated by this specific Target. It contains some metadata
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Remove the unused comments?

# within any given transitive closure, no namespace collisions
# exist. These MUST NOT begin with "/" or ".", or contain "..".
###modules=js.modules + modules,
# NestedSet<File> of all protobuf definitions in the transitive
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Remove the unused comments?

@@ -292,6 +290,34 @@ def compile_ts(ctx,
if not is_library:
files += depset(tsickle_externs)

rerooted_es6_sources = []
es6_js_modules_roots = [] # todo(achew22): this is non recursive... recurse.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@googlebot
Copy link

So there's good news and bad news.

👍 The good news is that everyone that needs to sign a CLA (the pull request submitter and all commit authors) have done so. Everything is all good there.

😕 The bad news is that it appears that one or more commits were authored by someone other than the pull request submitter. We need to confirm that all authors are ok with their commits being contributed to this project. Please have them confirm that here in the pull request.

Note to project maintainer: This is a terminal state, meaning the cla/google commit status will not change from this State. It's up to you to confirm consent of the commit author(s) and merge this pull request when appropriate.

@achew22
Copy link
Member Author

achew22 commented Nov 28, 2017

Hey @alexeagle. I think this works as is. Would you mind taking a look at it?

@mrmeku can you explicitly state you'd be okay with this PR being merged?

Copy link
Contributor

@mrmeku mrmeku left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Aside from where I commented I think this is ready for merging.

WORKSPACE Outdated
sha256 = "91fca9cf860a1476abdc185a5f675b641b60d3acf0596679a27b580af60bf19c",
url = "https://github.com/bazelbuild/rules_go/releases/download/0.7.0/rules_go-0.7.0.tar.gz",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: Consider reverting this part

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

FYI I have a PR out that upgrades this to 0.7.1

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

RM'd

name = "closure_compiled",
deps = [":parent"],
defs = [
"--jscomp_off=reportUnknownTypes",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we should preemptively change this to --jscomp_off=checkTypes . I think that flag disables all the flags that are going to bite us in the future as we expand this example more.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we need that? tsickle should suppress checkTypes on each JS file:
https://github.com/angular/tsickle/blob/8ebfd011e9386997ca94001210596c5fd75cd17d/test_files/default/default.js#L3

How do we deal with defs in general? I think internally at Google we've been learning towards wrapping with a macro like ts_binary to make things more uniform. But in open-source there is no uniformity.

/cc @evmar @rkirov

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We may need both. The generated output definitely has suppression on checkTypes, but without this it is failing to compile ATM

@@ -292,6 +291,25 @@ def compile_ts(ctx,
if not is_library:
files += depset(tsickle_externs)

rerooted_es6_sources = []
es6_js_modules_roots = []
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You can remove the entire part about es6_js_module_roots and this still compiles. I think you should remove it for now because we obviously don't quite understand what this does correctly. As far as what I've read from Jart commenting on github, these roots are just stripped off from es6 import statements. I think they might only be applicable for absolute import statements (since Bazel creates a new file system)

WORKSPACE Outdated
remote = "https://github.com/bazelbuild/rules_nodejs",
tag = "0.1.11",
commit = "f6eb823e1140da6eafbf77f56a249f72bb89416e",
remote = "https://github.com/achew22/rules_nodejs",
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this right? Should it be reverted before merging?

Copy link
Contributor

@alexeagle alexeagle left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

awesome work pushing this forward

WORKSPACE Outdated
remote = "https://github.com/bazelbuild/rules_nodejs",
tag = "0.1.11",
commit = "f6eb823e1140da6eafbf77f56a249f72bb89416e",
remote = "https://github.com/achew22/rules_nodejs",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you have a PR on rules_nodejs that has the bits you need? We can't have rules_typescript on master pointing to your fork of a dependency

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It doesn't actually matter now. That change was to allow 0.8.0rcN work when I was testing. now that 0.8.0 is out it shouldn't matter.

WORKSPACE Outdated
sha256 = "91fca9cf860a1476abdc185a5f675b641b60d3acf0596679a27b580af60bf19c",
url = "https://github.com/bazelbuild/rules_go/releases/download/0.7.0/rules_go-0.7.0.tar.gz",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

FYI I have a PR out that upgrades this to 0.7.1

name = "closure_compiled",
deps = [":parent"],
defs = [
"--jscomp_off=reportUnknownTypes",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do we need that? tsickle should suppress checkTypes on each JS file:
https://github.com/angular/tsickle/blob/8ebfd011e9386997ca94001210596c5fd75cd17d/test_files/default/default.js#L3

How do we deal with defs in general? I think internally at Google we've been learning towards wrapping with a macro like ts_binary to make things more uniform. But in open-source there is no uniformity.

/cc @evmar @rkirov

@@ -0,0 +1,3 @@
declare function someDeclaredFunction(s: string): string;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is there something in a test that asserts this isn't renamed?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Right now the example leaves the compiler in simple optimizations which shouldn't suffer from a renaming problem. Given that turning it on will require significant knowledge of JSCompiler I'd like to burn that bridge when we get there since it will involve using the compiled output in a end to end test to verify renaming.

executable=True,
cfg="host"),
"tsconfig": attr.label(
allow_files = True,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

curious, did you run buildifier? we should make that part of the process (maybe in a git pre-commit hook?)

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, I use the same .vimrc at home and at work and you can get buildifier with a relatively simple go install command. I don't even notice it any more

@@ -17,6 +17,7 @@

load(":common/module_mappings.bzl", "module_mappings_aspect")
load(":common/json_marshal.bzl", "json_marshal")
load("@io_bazel_rules_closure//closure/private:defs.bzl", "collect_js")
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

does this require every user of rules_typescript to add rules_closure to their WORKSPACE? I don't think that's desirable, as many users will choose rollup/uglify instead. How can we avoid that?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The only way I know to handle this is to copy/paste the code into this repo. You and I both gave the reviewer feedback to Dan asking him to just import it so there is a reflexive repulsion to both strategies.

@mrmeku you had a short implementation of this (<20 lines) could you paste it in here so I can drop it in the Change?

root = source.dirname
es6_js_modules_roots.append(root)

file_path = "%s/%s" % (root, source.basename.replace(".closure", ""))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

do we still produce the foo.closure.js outputs too?
I expected to see a change around https://github.com/bazelbuild/rules_typescript/blob/master/examples/es6_output/es6_output_test.sh#L5

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This copies the file to a new nested location so it shouldn't remove them. That is to say this test passes.

This looks to me like a fairly brittle test. Seems to me adding another one for this process is unnecessary and would be better verified by adding a closure_js_binary rule that actually consumes the output and emits a compiled .js (then possibly runs it through e2e with a browser).

achew22 and others added 3 commits November 30, 2017 23:00
Months of teath nashing later...
Every the ts_library impl now returns a struct named
"closure_js_library" needed by both closure_js_library and
closure_js_binary. This struct contains the srcs of the ts_library
transpiled to es6 as well as the path prefixes which should be stripped
off of import statements in order to find the imported file within the
file system.
@mrmeku
Copy link
Contributor

mrmeku commented Dec 20, 2017

You can close this now Andrew, I've done all this logic in a more Alex approved way in #112

@achew22
Copy link
Member Author

achew22 commented Dec 20, 2017

Perfect!

@achew22 achew22 closed this Dec 20, 2017
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

6 participants