> Byterun is a Python interpreter written in Python. This may strike you as odd, but it's no more odd than writing a C compiler in C.
I'm not so sure. The difference between a self-hosted compiler and a circular interpreter is that the compiler has a binary artifact that you can store.
With an interpreter, you still need some binary to run your interpreter, which will probably be CPython, making the new interpreter redundant. And if you add a language feature to the custom interpreter, and you want to use that feature in the interpreter itself, you need to run the whole chain at runtime: CPython -> Old Interpreter That Understand New Feature -> New Interpreter That Uses New Feature -> Target Program. And the chain only gets longer, each iteration exponentially slower.
Meanwhile with a self-hosted compiler, each iteration is "cached" in the form a compiled binary. The chain is only in the history of the binary, not part of the runtime.
---
Edit since this is now a top comment: I'm not complaining about the project! Interpreters are cool, and this is genuinely useful for learning and experimentation. It's also nice to demystify our tools.
gwerbin [3 hidden]5 mins ago
PyPy handled this by implementing PyPy in a restricted minimal subset of Python that they called RPython, and that seemed to work out well for them.
mikepurvis [3 hidden]5 mins ago
I was never a user of PyPy but I really appreciated the (successful) effort to cleanly extract from Python a layer that of essential primitives upon which the rest of the language's features and sugar could be implemented.
It's more than just what is syntax or a language feature, for example RPython provides nts classes, but only very limited multiple inheritance; all the MRO stuff is implemented using RPython for PyPy itself.
paulddraper [3 hidden]5 mins ago
The key difference is that RPython is actually a compiled language.
I.e. PyPy DOESN'T have an interpreter written in an interpreted language.
SJC_Hacker [3 hidden]5 mins ago
This is the case only if the new interpreter does not simply include the layer that the old interpreter has for translating bytecode to native instructions. Once you have that, you can simply bootstrap any new interpreters from previous ones. Even in the case of supporting new architectures, you can still work at the Python level to produce the necessary binary, although the initial build would have to be done on an already supported architechture.
direwolf20 [3 hidden]5 mins ago
Interpreters don't translate bytecode to native instructions.
SJC_Hacker [3 hidden]5 mins ago
The usual understanding of "interpreter" in a CS context is program that executes source code directly without a compilation step. However the binary that translates an intermediate bytecode to native machine code is at least sometimes called a "bytecode interpreter".
The article contrasts Python to Perl, saying Perl is purely interpreted while Python has compilation. This is factually incorrect.
Perl is transformed into an AST. Then that is decorated into an opcode tree. The thing runs code nearly as fast as C in many instances, once the startup has completed and the code is actually running.
anitil [3 hidden]5 mins ago
Oooh it's a bytecode interpreter! I was wondering how they'd fit a parser/tokenizer in 500 lines unless the first was `import tokenizer, parser`. And it looks like 1500ish lines according to tokei
I think because python is a stack-based interpreter this is a really great way to get some exposure to how it works if you're not too familiar with C. A nice project!
jgbuddy [3 hidden]5 mins ago
one liner:
eval(str)
PhunkyPhil [3 hidden]5 mins ago
I can do you one better:
```python3
from openai import OpenAI
import sys
client = OpenAI()
response = client.chat.completions.create(
model="gpt-4",
messages=[{
"role": "user",
"content": f"generate valid python byte code this program compiles to: {sys.argv[1]}"
}]
)
print(response.choices[0].message.content)
```
Actually, probably not better.
nasretdinov [3 hidden]5 mins ago
Went into comments looking for this exact comment. Wasn't disappointed
_blk [3 hidden]5 mins ago
Great minds think alike ;)
vachanmn123 [3 hidden]5 mins ago
Very well written! Everyone used to tell me during Uni that stacks are used for running programs, never ACTUALLY understood where or how.
And, in some ways, PyPy. I still think it is the sanest way to implement Python.
It makes me sad that I have to write C to make any meaningful changes to Python. Same goes for ruby. Rubinius was such a nice project.
Hacking on schemes and lisps made me realize how much more fun it is when the language is implemented in the language itself. It also makes sure you have the right abstractions for solving a bunch of real problems.
anitil [3 hidden]5 mins ago
> And, in some ways, PyPy
What do you mean by that? I'm not familiar with PyPy
The fact that it's written in python is often brought up in order to explain its name. But really, it's much less interesting than the fact that it has a tracing JIT. If it were called PyJIT I'd bet it would be clearer and more obvious that it's fast. And people would prob get less hung up on the distinction between python/rpython.
Doxin [3 hidden]5 mins ago
PyPy is python implemented in RPython, which is technically a python subset. It's so restricted it might as well be a different language though.
bjoli [3 hidden]5 mins ago
It is restricted in a way that you would restrict yourself to write high speed software in most languages, and I found it is not that restrictive compared to C that you would have to use if you were to write a fast Python library.
Doxin [3 hidden]5 mins ago
oh for sure, but I still feel like telling people pypy is written in python is misleading. it's written in something significantly like python, but it's not python.
mjmas [3 hidden]5 mins ago
> technically a python subset
So it can just run under CPython? If so, then that isn't too misleading.
bjoli [3 hidden]5 mins ago
Yes. It can run under Cpython (2.7).
actionfromafar [3 hidden]5 mins ago
Well, one could rewrite Python (perhaps piece by piece?) in Shedskin.
Shedskin is very nearly Python compatible, one could say it is an implementation of Python.
woadwarrior01 [3 hidden]5 mins ago
aka A Metacircular Interpreter
mapontosevenths [3 hidden]5 mins ago
Do you think God stays in heaven because he too lives in fear of what he's created?
blueybingo [3 hidden]5 mins ago
the article glosses over something worth pausing on: the `getattr` trick for dispatching instructions (replacing the big if-elif chain) is actaully a really elegant pattern that shows up in a lot of real interpreters and command dispatchers, not just toy ones -- worth studying that bit specifically if you're building anything with extensible command sets.
johndough [3 hidden]5 mins ago
Are you a bot? All your recent comments point out a thing in an article and contain LLM-isms.
HN.zip
I'm not so sure. The difference between a self-hosted compiler and a circular interpreter is that the compiler has a binary artifact that you can store.
With an interpreter, you still need some binary to run your interpreter, which will probably be CPython, making the new interpreter redundant. And if you add a language feature to the custom interpreter, and you want to use that feature in the interpreter itself, you need to run the whole chain at runtime: CPython -> Old Interpreter That Understand New Feature -> New Interpreter That Uses New Feature -> Target Program. And the chain only gets longer, each iteration exponentially slower.
Meanwhile with a self-hosted compiler, each iteration is "cached" in the form a compiled binary. The chain is only in the history of the binary, not part of the runtime.
---
Edit since this is now a top comment: I'm not complaining about the project! Interpreters are cool, and this is genuinely useful for learning and experimentation. It's also nice to demystify our tools.
It's more than just what is syntax or a language feature, for example RPython provides nts classes, but only very limited multiple inheritance; all the MRO stuff is implemented using RPython for PyPy itself.
I.e. PyPy DOESN'T have an interpreter written in an interpreted language.
https://doc.pypy.org/en/latest/interpreter.html
Perl is transformed into an AST. Then that is decorated into an opcode tree. The thing runs code nearly as fast as C in many instances, once the startup has completed and the code is actually running.
I think because python is a stack-based interpreter this is a really great way to get some exposure to how it works if you're not too familiar with C. A nice project!
eval(str)
```python3
from openai import OpenAI
import sys
client = OpenAI()
response = client.chat.completions.create( model="gpt-4", messages=[{ "role": "user", "content": f"generate valid python byte code this program compiles to: {sys.argv[1]}" }] )
print(response.choices[0].message.content)
```
Actually, probably not better.
It makes me sad that I have to write C to make any meaningful changes to Python. Same goes for ruby. Rubinius was such a nice project.
Hacking on schemes and lisps made me realize how much more fun it is when the language is implemented in the language itself. It also makes sure you have the right abstractions for solving a bunch of real problems.
What do you mean by that? I'm not familiar with PyPy
It lags behind CPython in features and currently only supports Python versions up to 3.11. There was a big discussion a month ago: https://news.ycombinator.com/item?id=47293415
But you can help! https://pypy.org/howtohelp.html
https://opencollective.com/pypy
So it can just run under CPython? If so, then that isn't too misleading.
Shedskin is very nearly Python compatible, one could say it is an implementation of Python.
cf: https://www.cs.cmu.edu/~rdriley/487/papers/Thompson_1984_Ref...
the text is based on python 3.5 which was released in 2015
other discussions:
https://news.ycombinator.com/item?id=16795049
https://news.ycombinator.com/item?id=12455104
https://news.ycombinator.com/item?id=11796253