Hacker News Reader

Comment by rayiner | Hacker News Reader

For someone who doesn't know either, but is familiar with programming generally, which is more readable?

I can guess what 'with-lines-from-file' does, I can guess what 'string-split' does, i can guess what 'string=' does, etc. The only thing I might not understand is 'aref'.

With the AWK example, I have no place to start. Symbol-oriented languages, rather than word-oriented languages, are like that. They tend to be more concise for things very specialized tasks, like this, but that limits their generality. Show me a matrix multiplication routine in AWK versus Common Lisp.

View on HN

I think you're agonizingly close to the point the author is quoting Peter da Silva as trying to make in your last paragraph here. Roughly sketched -

1) text and file processing are very common tasks 2) Lisp is not especially suited at these tasks, though it is very well suited for some other tasks. 3) If a dialect of lisp more suited to these very common tasks were made, it would increase the popularity of Common Lisp for the tasks that it works well for by familiarizing people with it.

It is both strictly correct and irrelevant to the topic at hand that the Lisp is easier to read. Because that's like saying "Lisp sucks because WTF does 'cddadr' mean anyway?"

AWK is a purpose-built language for splitting lines into fields and doing stuff with the result, so of course there's not a lot of boilerplate for splitting lines into fields and doing stuff with the results.

Lisp for doing a lot of fancy work with s-expressions is going to be high density code that's a lot harder for newcomers to read than verbose, plodding Java code that is working with a simple LinkedList class or something.

Readability (Edit: to newcomers) just isn't the relevant benchmark here.

I don't think a linked list in Java is a good example. I took two CS intro courses in a row last year; the first one used Scheme and the second one concentrated on data structures and used Java. There were some completely inexperienced people taking the two courses along with me.

As far as I could tell they had much more difficulty with linked lists and trees in Java, which made them needlessly complex, than they did in Scheme. And this is with people who had actually learned about both concepts first in Scheme, so they had more experience when they came to Java.

Fair point. Perhaps I picked a bad example.

At the same time, if you were writing a bunch of random awk-like one-offs as part of your normal lisp workflow, you wouldn't write 4 lines of code every time, right? You sit down and write some functions or macros that condensed it into something simple that allowed you to easily specify things like a split token, one or more filter-on fields and values (or better, a filter lambda), and one or more output fields, I bet. Because it's silly to write several lines for something you do a lot.

And while you might call it process-file-and-return-matching-results or something, the interface to the result might not be so different from the line of awk in the original post.

Reply on news.ycombinator.com