Kevin Reid's blog

I just wrote my first use of that <canvas> thing that everyone's talking about. (If you don't know, this is part of the WHATWG/HTML5 general vicinity of features: an element which creates a bitmap pixmap image that can be drawn on by JavaScript.)

Since LJ doesn't let me embed JavaScript or iframes (a restriction which I'm only just noticing...) I'll have to provide only a link (but your CPU meters will thank me for that):

Brownian Tree

(This is also part of a personal goal to learn more about Modern Web Applications technologies: “Ajax”, jQuery, web frameworks, etc. I used to (a) not have a personal server, and (b) tend to “All web content must be accessible to non-JS and pure-text users” — but I’ve come to realize that while one should still avoid unnecessary dependence on fancy stuff, there is value in using these systems to create what are not so much “web pages” as cross-platform, zero-install applications — they may not be usable by $NON-MAINSTREAM-BROWSER-TYPE but nothing else would be more compatible and still provide the same benefits.)

One bit of advice sometimes given to the novice programmer is don't ever compare floating-point numbers for equality, the reason being that floating-point calculations are inexact, and one should use a small epsilon, allowable error, instead, e.g. if (abs(value - 1.0) < 0.0001).

This advice is actually wrong, or rather, overly strong. There is a situation in which it is 100% valid to compare floats, and that is an cache or anything else which is comparing a float with, not a specific constant (in which case the epsilon notion is appropriate), but rather a previous value from the same source; floating-point numbers may be approximations of exact arithmetic, but that doesn't mean you won't get the same result from the same inputs.

So, don't get any bright ideas about outlawing aFloat == anotherFloat.

Unfortunately, there's a case in which the common equality on floats isn't what you want for previous-value comparison anyway: for most definitions of ==, NaN ≠≠ NaN. This definition makes sense for numerics (and is conformant to IEEE floating point specifications), because NaN is “not a number”; it's an error marker, provided as an alternative to exceptions (or rather, floating point error signals/traps/whateveryoucallit) which propagates to the end of your calculation rather than aborting it and requiring immediate error handling, which can be advantageous in both code simplicity and efficiency. So if you think about calculating within the space of “numbers”, then NaN is outside of that. But if you're working in the space of “results of calculations”, then you probably want to see NaN == NaN, but that may not be what you get.

Mathematically, the floating-point comparison is not an equivalence relation, because it is not reflexive on NaN.

(It's also typically the case that 0 == -0, even though positive and negative zero are distinct values. Oh, and NaNs carry data, but I'm not talking about that.)

What to do about it, in a few languages:

JavaScript

Even the === operator does not compare identities rather than numeric values, so if you want to compare NaN you have to do it as a special case. Google Caja handles it this way:

/**
 * Are x and y not observably distinguishable?
 */
function identical(x, y) {
  if (x === y) {
    // 0 === -0, but they are not identical
    return x !== 0 || 1/x === 1/y;
  } else {
    // NaN !== NaN, but they are identical.
    // NaNs are the only non-reflexive value, i.e., if x !== x,
    // then x is a NaN.
    return x !== x && y !== y;
  }
}

Common Lisp

The = operator generally follows the IEEE comparison (if the implementation has NaN at all) and the eql operator does the identical-object comparison.

E

The == operator is guaranteed to be reflexive, and return false for distinguishable objects, so it is appropriate for the “cache-like” use cases, and the <=> operator does conventional !(NaN <=> NaN), 0.0 <=> -0.0 floating-point comparison.

One of the things I’ve been procrastinatingah, not had the time to do, being busy with school and other projects, is announcing and working on a job search for this summer. I have posted my resume, but I didn’t even get around to mentioning that. The process really doesn’t excite me that much — it’s essentially research, comparison shopping, which I have never been very fond of.

But, last October, I was contacted out of the blue by a recruiter asking if I was interested in opportunities at — Google. After checking that it wasn’t a spoof I naturally said yes, and after a number of rounds of information exchange and interviews,

This summer, I will be (well, subject to my completing the process of accepting the offer) working as a Software Engineering Intern at Google, with the Caja team, in Mountain View, CA.

So — whoa and yay and other such cheerful words. And thanks to my friends at Google who referred me and nudged the process along.^*

The most uncertain remaining step is finding housing in or near Mountain View (could be as far as San Francisco or San Jose; Google runs a shuttle bus and is convenient to public transportation). Google has provided some general advice-for-interns, but I’d like to hear input from my readers and friends who already live in in the area.

Some parameters:

• I would consider living with other people, but I wouldn’t want to take a chance on a complete unknown. (So if you are someone or know someone with a room...)
• Speaking of taking chances, make the chance of being mugged on the way home in the evening very small, please.
• I am traveling from the east coast, probably by train, so I don’t want to have to transport a lot of stuff, or buy items that I’ll use for only three months — so a furnished space is better.
• I do not own a car, but I know how to drive one.
• I do not own a bicycle, but I used to know how to ride one.
• This will be the first time I have lived outside of my home city for longer than a week’s visit/vacation.

^*Y’know how job search advice is big on saying you should be “networking”? If you’ve thought you’re too much of the non-face-to-face-social non-polite-small-talk would-rather-talk-to-people-through-the-computer sort for that — take me as an example. This opportunity came to me because of other people who knew me entirely through my work on open source projects (E, and thus Caja-CapTP) — I didn’t do anything that I wouldn’t have done for other reasons anyway. I’m not saying you shouldn't do any of the other stuff you might be thinking of — I’m saying this stuff counts.

import Char
import Debug.Trace
import System.Environment
import Control.Monad.Fix

{-
  Hvm.hs
  By Kevin Reid, 2008-06-15

  This is an implementation of the virtual machine described at 
  <http://www.hacker.org/hvm/>. It was designed to fit into a single IRC line
  so that its definition could be put into lambdabot
  <http://www.haskell.org/haskellwiki/Lambdabot>, and its inspiration and
  creation was documented at <http://swhack.com/logs/2008-06-15>.

  Haskell source not in HTML: <http://switchb.org/kpreid/2008/Hvm.hs>
  Non-blog location: <http://switchb.org/kpreid/2008/Hvm.html>

  Explanation of variables:

    r = VM execution function of call stack, memory, instruction pointer, and 
        operand stack
    c = call stack
    m = memory
    i = instruction pointer
    j = instruction pointer plus one
    o = operand stack

    a = arithmetic handler
    h = proceed with instruction pointer and operation stack modification
    g = proceed with operand stack modification

    z:x = one value popped from operand stack
    z:w:e = two values popped from operand stack
    k:l = one value popped from call stack

    d, t: slice lists using z as length

  To make it fit in IRC, discard comments, imports, and "main", replace 
  all line breaks with semicolons as appropriate, and discard the ' ' and _ 
  cases (i.e. the interpreter will not support whitespace or report invalid 
  characters).
    perl -0777 -pe 's/\A.*(?=^s)//ms; s/\s*main .*//; s/--.*$//mg;
                    s/\n */;/g; s/[;\s]*(where|let)[;\s]*/$1 /g;
                    s/;'\'' '\''->g o;_->error\$show\(p!!i\)//;'
-}

s p=r[](fix(0:))0[]
  where
   r c m i o=--trace (show (m,i,(if length p<=i then '!' else p !! i),(reverse o),c)) $
     let a(&)=g$w&z:e
         j=1+i
         h=r c m
         g=h j
         z:x=o
         w:e=x
         k:l=c
         d=drop(1+z)
         t=take z
      in case(p++"!")!!i of
         'p'->show z++g x
         'P'->chr(mod z 128):g x
         d|isDigit d->g$ord d-48:o
         '+'->a(+)
         '-'->a(-)
         '*'->a(*)
         '/'->a div
         ':'->g$signum(w-z):e
         'g'->h(z+j)x
         '?'->h(i+case w of 0->z;_->1)e
         'c'->r(j:c)m z x
         '$'->r l m k o
         '<'->g$m!!z:x
         '>'->r c(t m++w:d m)j e
         '^'->g$x!!z:x
         'v'->g$x!!z:t x++d x
         'd'->g x
         '!'->""
         ' '->g o
         _->error$show(p!!i)

main = do [p] <- getArgs; putStrLn $ s p

A machine I used to use to host some web services, bots, and repositories became no longer accessible from the Internet, as a result of which I've had to move what I was serving from it; some to switchb.org, some to personal machines.

I took the opportunity to clean things up a bit, as a result of which I now have better backups, more polished services, and know a little bit more about configuring Apache — though not as much as I perhaps should.

• My Subversion repositories are now served over HTTP, and therefore browsable; and they are now backed up daily (using svnsync triggered by a cron job) to my laptop, and thence to all its backups.

  (I wasted several minutes on remembering that cron will ignore the last line of a crontab file if it doesn't end with a newline; after listening to me grumbling about this, someone made a suggestion to end the file with a comment, so that the last line is harmless whether ignored or not, and also reminds one of the issue.)

  If you have a working copy of one of my repositories (E-on-CL, E-on-JavaScript, MudWalker, Den, etc.), here's a guide to the changed URLs.

• My other Tahoe-LAFS volunteer grid storage node is now residing on a machine on my home LAN.

• Finally, some simple data-querying web services I wrote for Waterpoint's word games have now been moved to switchb.org; I also took the time to prettify their URLs (no cgi-bin or .cgi) and write documentation.

I haven't yet gotten to working on the bots, darcs repositories, or miscellaneous other stuff I had there.

(Pondering moving my blog over to switchb.org as well so as to not have ads, especially now that I found I can still have LJ-friends by way of OpenID. (Hm, but reading friends-locked posts over RSS might not work since there's no username+password for LJ to accept. Anyone have experience with that situation?))

Apache configuration questions:

1. If I have multiple variants of a document (e.g. foo.html foo.pdf foo.txt) handled by MultiViews, so the canonical URL of the document is extensionless (“foo”), how do I properly control the default variant to serve in the event that the client does not express a preference via the Accept header? (Without doing so, I found that it would serve the .txt version, whereas I would prefer the HTML.) All that I found that worked was to create a type map file named “foo” with quality values, and force it to be interpreted as a type map using <FilesMatch>. This seems kludgy to me.
2. What is the right way to serve CGIs, not in a designated cgi-bin directory, and without any .cgi extension in the URL? I initially tried to apply mod_rewrite, but I couldn't get it to work such that /foo internally contacted foo.cgi whereas /foo.cgi redirected to /foo. I resorted to another <FilesMatch> explicitly listing the name of each CGI and doing SetHandler cgi-script.
3. What is the right way to handle “testing” vs. “deployment” configurations, where the relevant Directory, Location, etc may be different depending on which server or subdirectory the site is set up on? I see that one may use environment variables — should I just set up variables containing the path prefixes for the particular host before including the generic configuration file?

I was working on some automated document generation, building the process, revising the input document, and checking the results, and got tired enough of typing "make output/foo.txt && open output/foo.txt" and then going back and editing that line when I wanted to check the PDF version etc. that I wrote a tool which I hope will have more general application as well. I called it “maken”:

#!/bin/sh
# Make files and view the results.
make "$@" && open "$@"

Or it could be generalized into being a combinator, running any two commands, rather than just make and open, specified as the first two arguments:

#!/bin/sh
ca="$1"
shift
cb="$1"
shift
"$ca" "$@" && "$cb" "$@"

(Which, for the Haskell folks, should be called &&&, except that that would be annoying to enter in the shell.)

\documentclass{article}
\usepackage{verbatim}
\begin{document}
  \verbatiminput{\jobname.tex}
\end{document}

Both of the non-internet-based computer-related classes I'm taking this semester are (only offered as) once-per-week night classes, nominally ending at 9:30. So I'm talking with someone after class — — gee, it's bedtime already.

Java programming class (yes, I know Java already) today, introducing loops. Task: write a factorial program. I wrote mine to use int arithmetic if possible, floating-point if not. Test the edge cases, of course. It says:

500! ≈ Infinity

I'm glad I used the “approximately equals” sign.

A while ago I got the idea to write a “my experiences with JavaScript” post; unfortunately, I’ve forgotten what I was going to put in it. However, yesterday someone said approximately “I need to learn JavaScript in 3 hours, what should I know?”, and I thought of quite a few not-immediately-obvious things to point out.