Jan. 23rd, 2013

tim: Tim with short hair, smiling, wearing a black jacket over a white T-shirt (Default)
"So, it's meme time. Describe-what-you-do-using-the-most-common-thousand-words-in-American-school-fiction time." [personal profile] pseudomonas pointed out that the corpus used here is a bit weird, but constraints are fun, so I'm going to roll with it anyway.

People tell computers to do things by writing words. To make it easier, they come up with different "word sets" for the computer. There are word-sets that are built into computers, which we say are "low". And there are word-sets that people use to tell the computer what to do, which we say are "high". I work on one of the high word-sets.

One of the things that happens when people tell computers what to do is that people can get confused. Then, the computer does the wrong thing. When that happens, cars might not want to stop, or an up-goer could burst into fire. To make people less confused, a high word-set can have "types". A typed word-set doesn't let you put one sort of thing where a different sort of thing is supposed to go. We say a typed word-set is "safe" if someone showed that if your words use types the right way, then your words will do the thing they stand for and the computer won't get stuck trying to do it.

When people tell computers what to do, they usually want the computer to do it quickly. Some of the high word-sets are safe, but not so good for making computers go fast, because the words in them are very different from the low word-set that the computer uses. Other word-sets are very close to the low word-set, but they make it easier to get confused when you're writing words. The word-set I work on makes it easy to tell the computer to do things quickly, and also easy to be less confused while using it.

Finally, a computer you buy now is usually made of lots of little computers. It's hard to think about what all of the little computers should do at the same time, because you only have one brain to think with. One way to think about telling all the little computers to do is to stop them from sharing memory with each other. Instead, you can make them talk to each other by sending notes to each other. The word-set I work on lets you use this "note-passing" way of getting all the computers to do work at the same time.

How do we turn the words we write into things a computer can actually do? The answer is that we write more words to tell the computer how to turn words from our high word-set into words from the computer's low word-sets. Those words we write help the computer turn a few big words into a lot of small words. I work on one of those "computer-help things" for our high word-set. I fix parts of it where people got confused before, and sometimes I help change it to handle new and different words.


I'll just make one observation here: "computer" is in the corpus, but "language" isn't.
tim: Tim with short hair, smiling, wearing a black jacket over a white T-shirt (Default)
We're down to ten pending pull requests, which is the lowest it's been in a while!

It's hard to know what to write today, since it seems like I spent most of the day merging pull requests. Or trying to. Merging a pull request: if it looks OK from reading it on github, and it merges cleanly, I just click merge. If I'm worried, I pull it as a new branch and run the testsuite on my machine. Several times today, a test failed, which means someone didn't run the full testsuite (understandable since it can take an hour...) or else there was a conflict between their code and another recent change. I decided today I really wanted to clear out pull requests, so I would fix things myself instead of asking the submitter to fix the problem and submit a new pull request. I am particularly proud of merging this library fix, for which I had to read the comments several times just to figure out what was needed. (I blame sleep deprivation, not the commenters.)

Also, I worked a little bit on this bug relating to our syntactic sugar for for loops. The syntactic sugar is very nice, but in this case it leads to a confusing error message that takes some thought as to how to get straight (due to the baroqueness of our for construct).

Profile

tim: Tim with short hair, smiling, wearing a black jacket over a white T-shirt (Default)
Tim Chevalier

September 2014

S M T W T F S
 1234 56
78 910 111213
1415 16 17 18 1920
21 222324252627
282930    

Most Popular Tags

Style Credit

Expand Cut Tags

No cut tags