Instead of using core/snapshot_creator.rs, instead two crates are
introduced which allow building the snapshot during build.rs.
Rollup is removed and replaced with our own bundler. This removes
the Node build dependency. Modules in //js now use Deno-style imports
with file extensions, rather than Node style extensionless imports.
This improves incremental build time when changes are made to //js files
by about 40 seconds.
* Revert "port more ops to JSON (#2809)"
This reverts commit 137f33733d.
* Revert "port ops to JSON: compiler, errors, fetch, files (#2804)"
This reverts commit 79f82cf10e.
* Revert "Port rest of os ops to JSON (#2802)"
This reverts commit 5b2baa5c99.
Fixes some sed errors introduced in c43cfe.
Unfortunately moving libdeno required splitting build.rs into two parts,
one for cli and one for core.
I've also removed the arm64 build - it's complicating things at this
re-org and we're not even testing it. I need to swing back to it and get
tools/test.py running for it.
Resolves #1705
This PR adds the Deno APIs as a global namespace named `Deno`. For backwards
compatibility, the ability to `import * from "deno"` is preserved. I have tried
to convert every test and internal code the references the module to use the
namespace instead, but because I didn't break compatibility I am not sure.
On the REPL, `deno` no longer exists, replaced only with `Deno` to align with
the regular runtime.
The runtime type library includes both the namespace and module. This means it
duplicates the whole type information. When we remove the functionality from the
runtime, it will be a one line change to the library generator to remove the
module definition from the type library.
I marked a `TODO` in a couple places where to remove the `"deno"` module, but
there are additional places I know I didn't mark.
* Native ES modules
This is a major refactor of internal compiler.
Before: JS and TS both were sent through the typescript compiler where
their imports were parsed and handled. Both compiled to AMD JS and
finally sent to V8
Now: JS is sent directly into V8. TS is sent through the typescript
compiler, but tsc generates ES modules now instead of AMD. This
generated JS is then dumped into V8.
This should much faster for pure JS code. It may improve TS compilation
speed.
In the future this allows us to separate TS out of the runtime heap and
into its own dedicated snapshot. This will result in a smaller runtime
heap, and thus should be faster.
Some tests were unfortunately disabled to ease landing this patch:
1. compiler_tests.ts which I intend to bring back in later commits.
2. Some text_encoding_test.ts tests which made the file invalid utf8.
See PR for a discussion.
Also worth noting that this is necessary to support WASM
- Improves speed and binary size significantly.
- Makes deno_last_exception() output a JSON structure.
- Isolate::execute and Isolate::event_loop now return
structured, mapped JSError objects on errors.
- Removes libdeno functions:
libdeno.setGlobalErrorHandler()
libdeno.setPromiseRejectHandler()
libdeno.setPromiseErrorExaminer()
In collaboration with Ryan Dahl.
Refactors handlers.rs
The idea is that all Deno "ops" (aka bindings) should map onto
a Rust Future. By setting the "sync" flag in the Base message
users can determine if the future is executed immediately or put
on the event loop.
In the case of async futures, a promise is automatically created.
Errors are automatically forwarded and raised.
TODO:
- The file system ops in src/handler.rs are not using the thread pool
yet. This will be done in the future using tokio_threadpool::blocking.
That is, if you try to call them asynchronously, you will get a promise
and it will act asynchronous, but currently it will be blocking.
- Handlers in src/handler.rs returned boxed futures. This was to make
it easy while developing. We should try to remove this allocation.
- Adding a ModuleId type alias to specify original url or local file of
dependency
- Modifying ModuleMetaData class to contain ModuleId
- Adding a --deps flag
