Hacker News Reader

Comment by taeric | Hacker News Reader

This came from a thought I had after posting https://news.ycombinator.com/item?id=18312529

Any feedback is greatly welcome. Both in areas I have wrong in the programming, and in general presentation. Thanks!

Great post! A few observations.

- It seems like the difficulty level ramps up really quickly starting with the "So, what makes the lisp example different?" section. This was jarring enough that I pretty much glossed over it (too much effort required to process all the definitions at once).

As a more meta point, I wonder if symbolic differentiation is in fact a good example of highlighting lisp's differentiating features? One could write the same code in python/javascript with using lists and strings and recursion in a similar way, no?

Not that I have a better example to highlight lisp's differences, which I understand to be macros. I've read SICP and have written a fair amount of elisp and still have not once identified a good use case for writing a DSL nor even a macro for myself.

One thing I’ve done recently in Common Lisp with macros was write a suite of macros to make working with objective-c in Common Lisp less burdensome: with a couple reader macros, I can almost copy-paste objective-c code samples into my Common Lisp program and have them work.

Also, some of the biggest complaints about redux and other JavaScript libraries is the amount of boilerplate you have to write, leading to really complicated scaffolding and code generation utilities: Babel, for example, is essentially just a big macroexpander for JavaScript enabling, among other things, the use of language features as soon as they are standardized rather than having to wait years until a sufficient portion of the available browsers support the new feature.

> One could write the same code in python/javascript with using lists and strings and recursion in a similar way, no?

Then you're implementing a Lisp (just one with square instead of round brackets, and commas between list elements) within Python/JS.

Edit: to explain, you also need to either translate the result of the derivation back into Python/JS code (via stringification and eval or perhaps AST manipulation--i.e. a compiler of the list structures (Lisp) to Python/JS), or write an interpreter for the list structures (i.e. Lisp interpreter).

The homoiconicity is important. Python and ECMAScript (don't use "JavaScript", it's a trademark of Oracle's) are not homoiconic. Consider this Monte translation of OP's blog post: https://github.com/monte-language/typhon/blob/master/mast/fu... The main feature present in Lisps and in Monte but not in Python nor ECMAScript is quasiquotation. ES has template literals, and Python 3 has f-strings, but neither language can use them for pattern-matching, only value-building.

Yeah, I thought of saying more on the code I used from SICP, but don't feel i could add much to their treatment.

I think you could write something in other languages, but it would likely be incumbent to you to lex/tokenize the data. And then to convert back to a string to send back to eval.

You can also do something similar with annotation processing in Java. Or the macro languages in newer languages. Much more difficult to me, though.

For highlighting the features, I don't know of much. I like lisp because i find it fun. In large from the parsimony of the language. Though, i think it is layered in ways we often gloss over in that phrase.

> One could write the same code in python/javascript with using lists and strings and recursion in a similar way, no?

That point is made in the article. You'd need to invent a language for writing the formulas as strings, and handle it with a scanner, parser (for which you pick an output tree format). Then do the differentiation on the tree format, and finally reformat the transformed code back to a string to be passed to the character-based "eval".

Only tiny comment is that your syntax highlighting makes the names in function definitions almost impossible to read.

Thanks! I think it picked up the theme i had active at the time. Will pick a different one when back at that computer.

To circle back on this, I found that I have a bit of bad interaction between some overrides of the CSS I did and the "htmlization" of my buffers to make the output.

Regardless, did a quick rerun with a theme that should have been a bit more compatible. Will have to see about fully fixing up the general export some time. Again, thanks for the feedback!

Not really a feedback, but I wrote something very similar a bit more than two years ago: http://eskatrem.github.io/Newton-Raphson.html

That is a much more comprehensive treatment! For anyone that is curious on how you could inspect a program in python, I think that is a great article to compare the two. (Indeed, I didn't know the syntactic tree api.)

Reply on news.ycombinator.com