Coding Horror: Parsing Html The Cthulhu Way

http://www.codinghorror.com/blog/archives/001311.html

by Anon84 • 17 years ago

66 33 17 years ago

codinghorror.com

zyb09• 17 years ago

krakensden• 17 years ago

the title was good though

michael_dorfman• 17 years ago

The original (Zalgo'd) version of the StackOverflow answer is even funnier than Jeff's excerpt: http://stackoverflow.com/questions/1732348/regex-match-open-...

marbletiles3• 17 years ago

It is, but it doesn't answer the question. The asker doesn't ask how to parse HTML, he wants to match certain tag-like patterns, and not others. He doesn't define the document, or the context, and could plausibly be wanting to remove certain tags from a non-HTML text document or the like.

There are plenty of potential areas involving HTML tags where you can manipulate the document with regexes, so long as you're not actually trying to make a proper parsing job of it. But everyone was so keen to show that they know you don't use regexes to parse HTML, they blew the actual question away. The asker did try to point out that he didn't want to parse the HTML, but still the points kept on rolling up for an answer that ... told him not to try and parse the HTML.

jordanb4• 17 years ago

While I agree that using regexes for one-off jobs is not the kind of thing that would summon Cthulhu. I think it's still the wrong thing to do and you would still be better off using a parser.

I think, to sum up this whole drama, we have one group saying "you should use a parser" and the other group saying "but sometimes regexes aren't totally horrible, and this is just a quick job. I don't want to take the effort to use the proper tool."

The latter group sees themselves as being pretty reasonable. And perhaps they are. But the thing that exasperates the former group is that it is generally not quicker or easier to use regexps -- even for one-off jobs --- than using a nice parser like BS or lxml.

Those tools give you a object-view of the document which, as JGC demonstrated, allows you to express what you want to do concisely and easily, even (especially) when you're picking a few tags out of the document.

The reason why regex hacking on X/HTML documents remains popular isn't because it's 'easy' but because it doesn't require people knowing how to use the proper tools, nor does it require people to make the conceptual leap from the document as a long string of unstructured text to the document being a tree-structure of objects.

It's a classic "The Wrong Way" choice made by people who don't have a mechanical or conceptual understanding of the right way.

It's like this. Imagine you come upon someone banging in screws with a hammer. And laying next to them is a nice power drill with a Phillips bit. You say to them "why aren't you using the drill for that. It'd do a better job and be a hell of a lot easier too".

And the guy responds "This is a quick job. These screws don't need to hold very much weight, or for very long. If this was a serious piece of construction I would totally use the drill, but for my needs this is just easier."

Now, you might look at him with some incredulity. No way is his way "easier." It's much harder, on top of doing a worse job. The real reason why he's not using the drill is, quite obviously, he doesn't know how!

So you say to him "seriously, try the drill" and he keeps on insisting that his approach is "totally sufficient" (which it may well be, that's not the point) and that using the drill would be "too much work."

And then that's how we got here in this discussion.

brown9-2• 17 years ago

I don't think this blog post does the original SO answer credit at all.

nkurz• 17 years ago

Beautiful effect, but I had to look up Zalgo: http://encyclopediadramatica.com/Zalgo

MindTwister• 17 years ago

You know its a bad day when encyclopediadramatica.com is a valid reference.

drtse4• 17 years ago

See http://www.jgc.org/blog/2009/11/parsing-html-in-python-with.... and http://news.ycombinator.com/item?id=923775 . Btw, as someone that has sometimes parsed html using regular expressions (for small and localized html snippets extracted by bigger pages) and has dealt with the numerous patches needed everytime something changes (structural changes or random \n,\r,etc... added to the source html ), i definitely agree that a html/xhtml parser is the way to go. The resulting implementation will be cleaner and the typical issues will be adressed slightly changing the parsing code or fixing bugs in the html cleaning/validator section of the parser...

PHP-TROLL4• 17 years ago

This is total bullshit. This is assuming that you are parsing well formatted, modern HTML.

Anyone who has had to data mine HTML on a regular basis knows regular expressions are the only way to go because most web pages are clusterfucks of invalid, font-tag laden, id-less, ancient piles of shit.

Any XML parser I threw at the sites I was mining ran away crying with it's pants soaked with urine. So climb down your ivory tower, because XML parsers may be the "correct" way to get data out of HTML, but they are near worthless for most situations.

psadauskas• 17 years ago

Thats because you're using an XML parser to parse HTML. You should use an HTML parser, like hpricot or beautifulsoup.

PHP-TROLL4• 17 years ago

Yeah when I said XML parsers I was speaking generically. I tried many HTML and XML parsers as well has HTML sanitizers. They worked, maybe, 60% of the time. Hpricot regularly segfaulted, beautifulsoup didn't even begin to work. Even if I ran the HTML through tidy first it would fail to parse.

Most people are ignorant of how bad most of the HTML on the web is, it is simply unparseable.

spuz• 17 years ago

It's not total bullshit. Did you get to the end of the article before commenting on it? I think Jeff hit the nail on the head; sure regex is generally not very good a parsing HTML, but in certain situations it is the right tool for the job. You seem to have considered your options and chosen the right tool. The point of the article is that often you don't need to teach someone which tool they should be using, but how to recognise that there might be better solutions to the problem they're solving and how to go about finding that solution.

JulianMorrison• 17 years ago

It's actually really simple.

You can't parse HTML with regex.

You can parse text blobs with regex.

If your input looks like HTML but is really a text blob, you can parse it with regex.

btilly• 17 years ago

No, it is even simpler than that. Regular expressions are for matching patterns, not parsing. They don't parse HTML. They don't parse text blobs. They don't parse anything. They just match.

cia_plant• 17 years ago

Don't HTML parsers use regexes for their lexical scan phase?

statictype• 17 years ago

I suppose that depends entirely on how the parser was implemented.

You could write a scanner that doesn't use regexes. It could just read in characters one at a time and hard-code the logic for matching tokens.

On the other hand, I wouldn't be surprised if many of the standard html parsing libraries use regex libraries internally to match tokens.

If you used a scanner generator like lex, you would specify your scanner using regexes and lex would generate code that expands those regexes into an automata. I guess that also counts as using regexes, only they're compiled down at compile-time instead of at runtime.

thaumaturgy• 17 years ago

Nice, codinghorror has audio ads for Mastercard which cause Safari to hang and then finally crash.

I loooove it when a site justifies my use of AdBlock.

Coding Horror: Parsing Html The Cthulhu Way

Related Stories