The Emacs Problem

73 points by mofey 16 years ago | 40 comments

<<What would you rather do? Learn 16 different languages and frameworks in order to do "simple" log-file and configuration-file processing? Or just buckle down, learn Lisp, and have all of these problems go away forever?

It's a rhetorical question. The answer is patently obvious at this point: Lisp is evil, and you'd damned well better write all your code in C++ and XML and JavaScript and PL*SQL and CSS and XSLT and regular expressions and all those other God-fearing red-blooded manly patriotic all-American languages from now on. No more of this crazy Lisp talk, ya hear?>>

snprbob86 16 years ago |

Uuugh... XML... how I hate you. Everyone should need to learn Lisp, even if you never ever program it, simply to avoid developing some of these poisonous XML-flavored tools. He mentions Ant specifically. Let me tell you, as an unwilling Microsoft Build expert, XML should be NOWHERE NEAR your build system. He puts down Makefiles, but as someone who wrote a Makefile for the first time after years of Ant and MSBuild, it was like a breath of fresh air. The intentional lack of power is what makes it so great. If anything, Makefiles are too powerful and too close to turing complete.

Here is what a make file should look like:

1) vpath statements for various file extensions

2) Phony targets for logical build units, including "all"

3) Generic rules for mapping .abc to .xyz; these rules should have exactly one line of code which executes an external script/tool

4) A list of edges in the dependency graph

5) A few variables as necessary to eliminate redundant edges

If you put any logic in your Makefiles, you are doing it wrong.

If your builds are slow, add empty dummy files for larger culling by timestamps. If timestamps are insufficient, codify early out logic into tools.

Not having logic in my Makefiles enables parallel execution and strong support for incremental builds. If I were to use Lisp as a build system, I'd create a DSL that had these same properties; forbidding arbitrary logic in my dependency graph. It's about finding the right balance to inject expressiveness without losing desirable properties of the DSL. This is why every developer needs to understand more about programming language design. Anytime you create any type of file format, you are doing it. And anytime you are writing any type of file format, you are reverse engineering it. Understanding the intended separation of logic for Makefiles helps me write better Makefiles.

1amzave 16 years ago | |

> If anything, Makefiles are too powerful and too close to turing complete.

"Close to"? Not a formal proof of Turing-completeness (and may only work with GNU make, not sure), but...

  [me@host: ~]% cat fibo.mk
  dec = $(patsubst .%,%,$1)

  not = $(if $1,,.)

  lteq = $(if $1,$(if $(findstring $1,$2),.,),.)
  gteq = $(if $2,$(if $(findstring $2,$1),.,),.)
  eq = $(and $(call lteq,$1,$2),$(call gteq,$1,$2))
  lt = $(and $(call lteq,$1,$2),$(call not,$(call gteq,$1,$2)))

  add = $1$2
  sub = $(if $(call not,$2),$1,$(call sub,$(call dec,$1),$(call dec,$2)))

  fibo = $(if $(call lt,$1,..),$1,$(call add,$(call fibo,$(call dec,$1)),$(call fibo,$(call sub,$1,..))))

  numeral = $(words $(subst .,. ,$1))

  go = $(or $(info $(call numeral,$(call fibo,$1))),$(call go,.$1))

  _ := $(call go,)
  [me@host: ~]% make -f fibo.mk
  0
  1
  1
  2
  3
  5
  8
  13
  21
  34
  55
  89
  144
  233
  377
  610
  987
  1597
  ^C

mahmud 16 years ago | | |

"The language of GNU make is indeed functional, complete with combinators (map and filter), applications and anonymous abstractions. GNU make does support lambda-abstractions."

http://okmij.org/ftp/Computation/#Makefile-functional

eklitzke 16 years ago |

One solution that Steve didn't discuss is JSON. To be fair, JSON wasn't that popular in 2005, but it's still a great solution.

The way it works is that their are no mandatory newline characters in JSON. Whitespace between lexical elements is ignored, and any embedded newlines in strings can be escaped (i.e. as \n). So a log format that a few people are using today is like this:

{'kind': 'foo', 'id': 1, 'msg': 'hi'} {'kind': 'bar', 'id': 2, 'msg': 'there'}

Each log message takes up a single line in the file. You can trivially deserialize any line to a real data structure in your language of choice. You can (mostly) grep the lines, and they're human readable. I do this at work, and frequently have scripts like this:

scribereader foo | grep 'some expression' | python -c 'code here'

In this case we're storing logs in the format described above (a single JSON message per line), and scribereader is something that groks how scribe stores log files and outputs to stdout. The grep expression doesn't really understand JSON, but it catches all of the lines that I actually want to examine, and the false positive rate is very low (<0.1% typically). The final part of the pipe is some more complex python expression that actually introspects the data it's getting to do more filtering. You can of course substitute ruby, perl, etc. in place of the python expression.

I feel like this is a pretty good compromise between greppability, human readability, and the ability to programatically manipulate log data.

panic 16 years ago |

You could even trivially convert it to XML and use XSLT, if you were silly enough. But Lisp is directly executable, so you could simply make the tag names functions that automatically transform themselves. It'd be a lot easier than using XSLT, and less than a tenth the size.

And now anyone who can modify your log can execute arbitrary code in the reader process…

ynniv 16 years ago | |

People always say that. Fortunately someone figured that out on the order of 34 years ago

[ http://en.wikipedia.org/wiki/Capability-based_security ]

(for a more recently active discussion, try http://en.wikipedia.org/wiki/Domain_Specific_Language)

Xurinos 16 years ago |

Did anyone else think of CL-PPCRE as they read through the "lisp does not have regular expression support" implications? That was answered by the comparison of elisp to modern Common Lisp, and I wonder if anyone has done any work to make CL-PPCRE work for elisp. In spite of being someone's library, it is much faster than perl's built-in regex support.

There is a point to be made for the idea that you are solving the wrong problem with the log parsing. On the other hand, if you are trying to interface with other developers' popular engines, you may not have a choice.

jrockway 16 years ago | |

Most of those benchmarks are against Perl 5.8, the version of Perl released in 2002. 5.10 had major regexp engine improvements, and 5.12 had minor improvements. Anything is fast when you compare it to 10 years ago.

btilly 16 years ago | | |

Furthermore the 5.8 engine is pretty much the 5.6 engine, which was significantly slower than the 5.005 engine. Why the slow-down? Because 5.6 has logic to check when regular expressions are matching slowly, and then to add tracing when it is. This makes many of the exponential slowdowns that Mastering Regular Expressions describes be automatically caught and handled fairly quickly.

The CL-PCRE and Java regular expression engines don't have those somewhat expensive checks, and so are much more likely to encounter catastrophic behavior.

Xurinos 15 years ago | | |

I actually ran my own tests to see the difference, including tests against hash speed and so forth. But it is good to know there are other benchmarks out there that correlate with what I found, even if older ones. Has anyone published new ones?

pragmatic 15 years ago |

Just to point out LINQ to XML http://msdn.microsoft.com/en-us/library/bb387098.aspx.

It takes a lot of pain out of XML processing. Don't have to remember the specifics of XPath/XQuery but you still have to deal with the pain of multiple namespace resolution inherit in XML.

jcapote 16 years ago |

It would be interesting to see an emacs implementation in language better suited for text manipulation like Ruby or Perl.

jrockway 16 years ago | |

I've written a lot of Emacs Lisp and a lot of Perl. Perl is better for bulk text processing, but that's not what editors do. The Emacs API is great for text editing. Once you are willing to treat blocks of text as buffers, everything gets very easy with Emacs Lisp. But most people are afraid to do that, because they are used to working with big opaque text objects and big batch transforms.

(Say you want to prefix every sentence with the expression "And then he said, ". The perl way: replace every empty space after a period or the beginning of the string with "And then he said, ". Emacs way: While there are sentences, go to the beginning of the next sentence and insert "And then he said, ". Same result, different way of thinking.)

Modify your way of thinking, and the Emacs model is wonderful. (Why do you think there are so many more Emacs extensions than Eclipse extensions, even though you can pretty much use any JVM langauge to customize Eclipse? It's because customizing Emacs is fast and easy.)

jgg 16 years ago | |

Hell, even Scheme would be a lot nicer than Emacs Lisp. Actually, if I could even get Yi ( http://www.haskell.org/haskellwiki/Yi ) to work, I'd use that. Elisp sucks.

jrockway 16 years ago | | |

So, what have you written in Elisp?

mechanician 16 years ago | |

As an emacs newbie with a nagging desire to learn Perl, I would gladly commit to switching to an emacs implementation in Perl.

jrockway 16 years ago | | |

Not that I would use it, but there is Padre:

http://padre.perlide.org/

angusgr 16 years ago | | |

If you think emacs lisp is hard to read now... ;)

obneq 16 years ago | |

Perfect Emacs Reimplementation Language

sammyo 16 years ago | | |

Another Reimplementation Conundrum?

lelele 16 years ago | |

Emacs Lisp is better suited for text manipulation than Perl or Ruby. Powerful regexes just don't cut it, you need text related abstractions.

Emacs clones based on both Scheme and Common Lisp have been written, but they haven't taken off.

vtail 16 years ago |

TextMate/Ruby = Emacs/Elisp

ynniv 16 years ago | |

But Emacs is capable of intelligently editing source code which is not Elisp. As much as I have tried, I cannot say the same of TextMate.

(How hard can it be to select blocks on double-click? Even Terminal.app does it!)

[me@host: ~]% cat fibo.mk dec = $(patsubst .%,%,$1) not = $(if $1,,.) lteq = $(if $1,$(if $(findstring $1,$2),.,),.) gteq = $(if $2,$(if $(findstring $2,$1),.,),.) eq = $(and $(call lteq,$1,$2),$(call gteq,$1,$2)) lt = $(and $(call lteq,$1,$2),$(call not,$(call gteq,$1,$2))) add = $1$2 sub = $(if $(call not,$2),$1,$(call sub,$(call dec,$1),$(call dec,$2))) fibo = $(if $(call lt,$1,..),$1,$(call add,$(call fibo,$(call dec,$1)),$(call fibo,$(call sub,$1,..)))) numeral = $(words $(subst .,. ,$1)) go = $(or $(info $(call numeral,$(call fibo,$1))),$(call go,.$1)) _ := $(call go,) [me@host: ~]% make -f fibo.mk 0 1 1 2 3 5 8 13 21 34 55 89 144 233 377 610 987 1597 ^C