My 8 hour journey to a single character(lbrandy.com)
lbrandy.com
My 8 hour journey to a single character
http://lbrandy.com/blog/2009/11/the-8-hour-journey-to-a-single-character/
10 comments
Telling expert humans what you're trying to achieve
This is critical. All sorts of "dumb questions" turn out to have been created by faulty assumptions or bad solutions to a larger problem. Often it's like traveling to the top of a mountain to boil water instead of fixing the broken stove. When asking for help, users should always ask the big picture question ("I need help setting up a server to do X, I'm getting error message Y when doing Z" rather than just the specific problem "I'm getting error message Y when doing Z").
This is critical. All sorts of "dumb questions" turn out to have been created by faulty assumptions or bad solutions to a larger problem. Often it's like traveling to the top of a mountain to boil water instead of fixing the broken stove. When asking for help, users should always ask the big picture question ("I need help setting up a server to do X, I'm getting error message Y when doing Z" rather than just the specific problem "I'm getting error message Y when doing Z").
An excuse to share my favorite bash.org quote:
<glyph> For example - if you came in here asking "how do I use a jackhammer" we might ask "why do you need to use a jackhammer" <glyph> If the answer to the latter question is "to knock my grandmother's head off to let out the evil spirits that gave her cancer", then maybe the problem is actually unrelated to jackhammers
http://bash.org/?866112
<glyph> For example - if you came in here asking "how do I use a jackhammer" we might ask "why do you need to use a jackhammer" <glyph> If the answer to the latter question is "to knock my grandmother's head off to let out the evil spirits that gave her cancer", then maybe the problem is actually unrelated to jackhammers
http://bash.org/?866112
As a corollary: when providing help, always demand the big picture. Frequently you will catch flak for this, in the form of "I am the expert in X, I know what I'm doing just fix the miniscule thing", but it is always great to see the look on someone's face when you say "Oh, dept. Y already does that and makes the result available at foo, it should save you a few hours of work".
I can personally attest that in the past I have spent an hour googling on a problem to figure out what to google to get a solution.
Sounds like a classic: How to Ask Questions the Smart Way
http://catb.org/~esr/faqs/smart-questions.html
http://catb.org/~esr/faqs/smart-questions.html
Good point. This is why in Extreme Programming (and Cucumber), user stories are supposed to include a "Why?".
"As a user, I should be able to mark my favorite posts, so that I can find them again." It may turn out that users don't need to find old posts, or would prefer to search instead of looking through an old list.
"As a user, I should be able to mark my favorite posts, so that I can find them again." It may turn out that users don't need to find old posts, or would prefer to search instead of looking through an old list.
> Telling expert humans what you're trying to achieve is far more useful.
Tiny little problem. You may not be able to.
Having worked in commercial, academic and government contexts, I can tell you that real life projects will always, always have some kind of hurdle to prevent you from divulging "what you're trying to do."
You might be working from a spec that only tells you what your superiors what you to know, there might be some legal or "market positioning" reason that requires you to preserve secrecy, or there might be department politics that will cause fur to fly if somebody outside your team finds out what you are really doing...
I sure hope my current stab at self-employment works out; I'd really had to go back...
Tiny little problem. You may not be able to.
Having worked in commercial, academic and government contexts, I can tell you that real life projects will always, always have some kind of hurdle to prevent you from divulging "what you're trying to do."
You might be working from a spec that only tells you what your superiors what you to know, there might be some legal or "market positioning" reason that requires you to preserve secrecy, or there might be department politics that will cause fur to fly if somebody outside your team finds out what you are really doing...
I sure hope my current stab at self-employment works out; I'd really had to go back...
It can sometimes be really hard to even communicate assuming you do have the ability to tell them everything they need to know about. I'm not sure the original poster would have been able to figure out what he should have done had he not gone down all these rabbit holes; you need to understand the true screwed-ness of visual formats to really grok it.
Any question I've ever needed to ask about $WORK_CODE has been abstract enough to ask without revealing any internal information.
On the other hand, one may argue that lbrandy is now an expert on color space conversions. He already had enough expertise to solve the problem and he now knows even more. Asking an expert wouldn't have helped him in achieving that.
But he did ask a (relatively) expert - his video developer friend.`
Exactly. I think all the people replying with their stories of missing braces or semicolons or typos are missing the point. This wasn't an issue that could've been solved by careful (character by character) reading. It was one that required a ton of research. Tracking down a missing brace doesn't teach you anything. Tracking down an issue like this does.
Agreed. I'll add that all the time he spent hunting down the bug put him into a position to really understand the question he was asking the expert.
As frustrating as these little problems can be, really digging in and taking the time to understand it is what experience is all about.
As frustrating as these little problems can be, really digging in and taking the time to understand it is what experience is all about.
I had the pleasure once of debugging an issue in C# that boiled down to the difference between a property "Percent" which was represented as e.g. 45.6, and a field "percent" which was measured as a fraction, e.g. 0.456. After correcting a reference to the field to be a reference to the property, I was proud to claim that my fix was a fix of a single bit - the difference between ASCII 'P' and ASCII 'p'.
Can't get smaller than that!
Can't get smaller than that!
I was at a pompously titled Software Craftsmanship conference, and one of the questions going around was: "what is the minimum commit size?"
My first guess was one character, since you can fix a bug by changing one character. But in the 'P' vs 'p' case, minimum size could be larger - fixing the calling code is fine, but is there a change that can be made to stop this bug appearing in other calling code? Eg, simplify the interface so that one of the 'p's is protected, so that the wrong Percent can't be set by accident.
My first guess was one character, since you can fix a bug by changing one character. But in the 'P' vs 'p' case, minimum size could be larger - fixing the calling code is fine, but is there a change that can be made to stop this bug appearing in other calling code? Eg, simplify the interface so that one of the 'p's is protected, so that the wrong Percent can't be set by accident.
One character. One friggin’ “J”. 8 hours. I hope no one is keeping track of “lines of code per hour”.
Sounds like he / you acquired some essential domain knowledge in those eight hours. Your code is more efficient, and you understand the problem space. That is what programming is about, not how many lines of code you write.
(When working on a large existing code-base, my goal is to remove lines of code from the project. Writing code means you are writing bugs. Removing code means you are removing bugs :)
Sounds like he / you acquired some essential domain knowledge in those eight hours. Your code is more efficient, and you understand the problem space. That is what programming is about, not how many lines of code you write.
(When working on a large existing code-base, my goal is to remove lines of code from the project. Writing code means you are writing bugs. Removing code means you are removing bugs :)
Whoosh.
Can you explain what apparently obvious point you think jrockway has missed?
jrockway bromidically responded to a throwaway joke that in itself already acknowledged the content of the response.
Judging from various codebases I've had to maintain, most programmers don't think this is a joke. They really think this guy wasted 8 hours learning how his code is supposed to work. They would have just reverted to last night's version and moved on to adding some more technical debt. (It's OK because it's test-driven development. If the tests pass, the code is fine!)
I'm sure most of the people here feel the same way. It just seemed like an audience mismatch, and if you acknowledged it (either the joke nature or the audience mismatch), I wouldn't have felt the need to "Whoosh" you.
I agree with your point but the TDD comment tacked on at the end is wrong and adds nothing.
I don't think I've ever heard anyone speak in support of TDD say that the code is fine if the tests pass. The canonical process of TDD explicitly calls for refactoring after the tests pass. That step is crucial and lasts as long as it takes for the programmer to be satisfied with the code (e.g. 8 hours).
I also found no evidence in the article that there was TDD involved. Perhaps I've misunderstood either the article or your comment. If that is the case I'd gladly retract my criticism.
I don't think I've ever heard anyone speak in support of TDD say that the code is fine if the tests pass. The canonical process of TDD explicitly calls for refactoring after the tests pass. That step is crucial and lasts as long as it takes for the programmer to be satisfied with the code (e.g. 8 hours).
I also found no evidence in the article that there was TDD involved. Perhaps I've misunderstood either the article or your comment. If that is the case I'd gladly retract my criticism.
If you google "bromidically", this response is now number 2 in Google - after dictionary.com.
I had a similar case; a regression that lasted 3 weeks and took half a day of grovelling through code to fix... and turned out to be two one-character typos in a #define name:
http://git.videolan.org/?p=x264.git;a=commitdiff;h=2389de25f...
http://git.videolan.org/?p=x264.git;a=commitdiff;h=2389de25f...
I just fixed a "should be || was &&" logic error that manifested itself fairly mysteriously in a one-dimensional partitioning data structure I had made(meant for indexing things like collision rectangles that have a "canonical value" and a "represented value" which need to be synchronized).
When it removed elements, it was matching both the represented value(an integer) and the node(an object reference). Using && meant that it removed too greedily. Because updates always went "remove then add," this error went undetected until I hit the case where two or more objects shared an identical reprval. Then it removed all of the shared objects, leaving only one behind when I went to query the partition. After going through "Does my query function work right, does my list implementation work right, " I finally narrowed it down to the add/delete functions.
When it removed elements, it was matching both the represented value(an integer) and the node(an object reference). Using && meant that it removed too greedily. Because updates always went "remove then add," this error went undetected until I hit the case where two or more objects shared an identical reprval. Then it removed all of the shared objects, leaving only one behind when I went to query the partition. After going through "Does my query function work right, does my list implementation work right, " I finally narrowed it down to the add/delete functions.
That sounds like an error that would have been detected by a unit test.
Being a little snide... that sounds like a unit test that occurs to you after encountering that bug.
It's a conceptual error with the problem, which means that it indeed WAS missed by unit tests. I wrote some of those and checked the "overlapping numbers" case; the difference was that I did so in a naive situation where the values were inserted once and never updated.
Now, it would have gone from a "mysterious" error to an obvious one had I done an academic-style step-by-step visualization applet of the entire structure's processes before integration, but I had confidence when the code was first written that because this structure was being used with extremely high frequency, the only thing I had to write an explicit test for off the bat was off-by-one ranges causing edge-case "near misses"; anything else would make itself known at runtime after integration.
Time it would have taken for a visualization applet: ~3-6 hours? Time it took to solve the bug: ~3 hours
It's placing a bet - and the bet is that the bugs are limited to a specific segment of code that can be narrowed down easily; if the problem were architectural in nature I'd be in far deeper shit because that would make the same class of bug appear all over, with any number of different symptoms.
Now, it would have gone from a "mysterious" error to an obvious one had I done an academic-style step-by-step visualization applet of the entire structure's processes before integration, but I had confidence when the code was first written that because this structure was being used with extremely high frequency, the only thing I had to write an explicit test for off the bat was off-by-one ranges causing edge-case "near misses"; anything else would make itself known at runtime after integration.
Time it would have taken for a visualization applet: ~3-6 hours? Time it took to solve the bug: ~3 hours
It's placing a bet - and the bet is that the bugs are limited to a specific segment of code that can be narrowed down easily; if the problem were architectural in nature I'd be in far deeper shit because that would make the same class of bug appear all over, with any number of different symptoms.
I love stories from the trenches like this! I'm currently diving into the shallow end of Lucene trying to accommodate ridiculous customer requirements. I'm sure in a couple hours there will be a "hidden flag" story somewhere in here.
The part of my code soul that dealt with search requirements goes out to you.
Something like this happened to me today. I noticed a floating menu that worked fine in Firefox and Safari didn't work in (wait for it) IE. After about three hours, I discovered that the JavaScript for the menu needed a CSS top attribute to work with; FF/Safari set it to 0 by default, but IE apparently needs an explicit value. Three hours, one line:
#navtool {
top: 0;
}
I now realize why people don't always share their failures in the way Feynman said we should in "Cargo Cult Science."
Comments like this:
> wow, you are an idiot. i feel bad for you. you work on image processing for a living?
Comments like this:
> wow, you are an idiot. i feel bad for you. you work on image processing for a living?
I would think there would be another cause of different results: chrominance being encoded at lower resolution, dynamically and spatially. So converting to RGB then to gray would make it more quantised and blurrier . . . well, I haven't thought this through in detail, so I am not certain, but it seems a possible consideration . . .
(since each channel of RGB is a weighted combination of all three YUV, each of RGB must be somewhat lower resolution, and that can't be recovered, so when converted to gray, that lower resolution must remain . . . so it seems it would be a deficiency . . . )
(since each channel of RGB is a weighted combination of all three YUV, each of RGB must be somewhat lower resolution, and that can't be recovered, so when converted to gray, that lower resolution must remain . . . so it seems it would be a deficiency . . . )
That's not lower resolution, that's lower precision. It's the difference between the number of samples and the accuracy of those samples.
This has nothing to do with luminance range though. Even if the output samples only had 4-bit precision, it wouldn't make black into gray.
This has nothing to do with luminance range though. Even if the output samples only had 4-bit precision, it wouldn't make black into gray.
I understand the difference, which is why I said 'lower resolution, spatially and dynamically'.
Are you saying definitively that chrominance is not encoded at lower spatial resolution?
I also said it might be another cause of difference, not the cause.
Anyone might want to look at http://en.wikipedia.org/wiki/4:2:2 before simply assuming I am talking nonsense.
Are you saying definitively that chrominance is not encoded at lower spatial resolution?
I also said it might be another cause of difference, not the cause.
Anyone might want to look at http://en.wikipedia.org/wiki/4:2:2 before simply assuming I am talking nonsense.
A little off-topic, but I was under the impression that using the ffmpeg commercially was problematic, did you find you had to address that issue at all?
ffmpeg is LGPL; it is used by hundreds of commercial applications around the world. For that matter, it has no commercial competitors for many of its use-cases.
Even On2, a competing encoder company, includes instructions with its encoder on how to set up mplayer (which uses ffmpeg) for decoding input videos.
Even On2, a competing encoder company, includes instructions with its encoder on how to set up mplayer (which uses ffmpeg) for decoding input videos.
Besides which, it sounds like they're only using it in-house for processing, not distributing anything based on it. So even if they're not compiling it in LGPL mode, they're probably perfectly OK.
The main 'problem' is that you'll still require an MPEG licence for encoding and decoding. Once you've bought one and abide by the usual LGPL rules (don't claim ownership, make source & changes available if distributing) then there shouldn't be any more commercial issues than, say, using Linux.
Googling is great, but it only helps when you know what you're looking for. Telling expert humans what you're trying to achieve is far more useful.