On 06/21/2016 03:39 PM, Christian Gollwitzer wrote:
Am 21.06.16 um 14:10 schrieb Tal Zion:
develop frontends in Java, Swift, Javascript, etc.
>
So how does this magic work? We developed a new compiler platform called
Bridge. At the heart of Bridge is the Bridge Extensible Code
Representation (BECR). Code in any language is parsed into an AST and is
then translated to the BECR. [.....]
easy as it**merely translates syntax. Then, Bridge translates the BECR
to LLVM IR and from there the code can be compiled to x86, arm, asm.js
or any other LLVM backend*.
So you have repeated .NET, or Parrot, or the JVM. What is new in your
system that sets it apart from these?
Soon Bridge will be released as an open source project. We’d love to
hear your opinions about this project. We’re really excited and we can’t
wait to see Python being used in more and more places!
Make sure you can do better than those:
http://ironpython.net/ (Python on .NET)
http://parrot.org/ (Python on the Perl-VM)
http://www.jython.org/ (Python on the JVM)
All of those projects promise a seamless integration of several
languages in a single VM. ALl of them have failed in the sense that
the languages have not converged onto that platform.
Christian
Hi Christian,
Thank you for your opinion, it is much appreciated.
There are similarities but there are also a few important implementation
differences:
* It is easier to add languages to Bridge than it is to any of the above
platforms: In order to make Python run on those platforms, those
projects needed to write a compiler from Python to those platforms'
respective intermediate languages. That is a very hard task. In Bridge,
we provide the BECR, which makes it easy to re-use concepts implemented
in other languages. For example, after we are done adding enough
concepts to the BECR to compile 100% of Python's features, adding Ruby
will be relatively easy as there are many similar concepts between the
two languages.
* Bridge enables Python code to interact with more languages through our
external interfaces: Bridge code can interact with libraries not
compiled to Bridge. So Python code will be able to call Swift,
Objective-C, Java, and C code, to name a few, even if they weren't
compiled to Bridge.
* Bridge integrates with CPython in order to provide the same experience
developers are used to: There are many Python modules written in C which
are not usable in the above platforms. In Bridge you can use any CPython
module. So when you are creating a Python list in Bridge, you are
actually using CPython's list implementation.
* Bridge makes Python faster: Python code compiled through Bridge is
compiled to native code. Because we are leveraging LLVM's many
optimizations, Python code will run faster than ever.
* Bridge is more hackable: Using plugins, Bridge is much more accessible
to community participation.
Most importantly, Bridge will enable Python code to run on many new
platforms (iOS, Android, web browsers..), while none of the above
platforms provide this ability. We aren't looking to compete with
CPython, but rather to complement it.
Tal
--
https://mail.python.org/mailman/listinfo/python-list
