I was at class when logins were launched...definitely registered the first day, though. It's a bit humbling to know that my login is almost old enough to drive.
Don't kid yourself -- there's a huge over-supply of graduate-level biologists with mathematical and computational training. The problem has nothing to do with the "softness" of the discipline, but rather, with the fact that academia pumps out doctorates at rates that can't be supported by industry.
Moreover, the button-pushers of the biology world are actually in a sweet spot, with regard to supply and demand. It's generally quite easy to obtain lab tech work with a BS (or even an AA). With one of these jobs, you'll live comfortably, work normal hours, and though you'll likely never lead a project, you'll earn respect and authority over time. An MS will bump you to a slightly higher salary, but it's questionable whether the gain is worth the opportunity cost (lost income, mainly).
By contrast, a PhD will leave you largely unemployable. You won't even be considered for most "PhD-level" positions, until you've completed an additional 2-4 years of post-doctoral "training", on top of the 5-7 years it takes to get the degree. And ironically, you won't even be able to get the lab tech work that you could have found with a BS/MS, because it is perceived as a "waste of talent" to put a PhD in a tech position, and most companies are fearful that you'll leave for greener pastures at the first opportunity (to be fair, this is probably true).
The OP is correct: if you're intelligent, and you're concerned about income, the biological sciences are a terrible place to be. Computer science is much more lucrative -- as are law, medicine, business and engineering. The perception that bio-tech is a job generator is largely a function of industry propaganda, and does not stand up to even casual scrutiny.
If you doubt me, go to the Science Magazine Careers Forum, and check out the number of truly sad stories in the field....
Have you ever actually *ridden* on a train in Japan? Outside of the major cities, it's pretty rural -- far more rural than anything but the most desolate parts of the American west. And ironically enough, it's actually *easier* to get higher speeds in rural areas, because there are fewer people to annoy, intersections to block, and so on.
Population density has nothing to do with the lack of efficient rail transit in the United States. We simply don't invest in it, because we're too busy subsidizing the air, car and trucking industries.
Well said, AC. I am perpetually annoyed by the number of people who bash on C++ simply because they don't want to take the time to learn a language that is more complicated than the ones they already know. There is more to the programming universe than Java and Perl!
Some goon gets a +5 moderation for admitting that the doesn't know C++ well enough to understand the compile errors he obtains, but the posts calling him out on his ignorance are met with "Troll" moderation. Welcome to slashdot.
"Complex syntax hinders communication between developers and increases defects. If everything else is equal (and it usually is), simpler syntax is better."
Granted, but the question remains: complex to whom?
I write template code (and use boost) on a regular basis, and I am familiar with the syntax. Once I got over the initial learning curve, I didn't find C++ template metaprogramming syntax to be particularly bizarre or difficult. I'm not trying to brag; I'm just pointing out that there is nothing inherently wrong with template code. It's understandable (and incredibly productive) if you take the time to learn to use the tool.
Again, this entire debate boils down to the question of what you find familiar and comfortable. If it were impossible for a programmer of normal intelligence to learn how to use boost, you might have a point; that's clearly not the case. In my experience, most complaints about C++ tend to be made by programmers who believe that their understanding of ALGOL-derived syntaxes should carry them comfortably through their programming careers. These types of people tend to get cranky and dismissive when they can't pick up a language in an afternoon.
"Fortunately, the next generation of programmers has many languages available that are not as complicated as C++. Progress in programming language design is indicated by syntactical simplicity, not complexity."
Demonstrably false. Progress in programming language (if it has been defined by a single metric at all), has been defined by increases in expressiveness, not simplicity. Cases in point: Java, Perl, Python and C are all significantly more syntactically complex (and expressive) than FORTRAN 77 and PASCAL.
"The GP sounds like he understands template metaprogramming very well."
Really? He was complaining about being unable to read the error messages.
"Nothing about his post suggests that he either a) says that he is an "expert" (as you are suggesting he does), or b) *isn't* enough of an expert to have formed a valid opinion on the subject."
You can't have it both ways: either the guy is informed enough to have an opinion on the matter, or he isn't. He made a post claiming that template metaprogramming is not "normal", while simultaneously making it absolutely clear that he didn't understand enough of what he was talking about to be even remotely qualified to advance a criticism.
Like it or not, there are quite a few C++ programmers out there who know how to work with complex templates. You may not know how, the GP poster may not know how, but that doesn't make the boost library "bad".
"It's only because of lots and lots of dedicated effort by Boost developers that Boost's very-complex template metaprogramming techniques work on a wide variety of popular compilers."
Straw-man. Any good library represents "lots and lots of dedicated effort" by developers, and almost by definition, wouldn't be good without that effort. Yet I doubt you're willing to abandon every domain-specific library, simply because you don't understand the code. Whether you realize it or not, your whole argument rests on that premise.
Keep in mind that the point of the Boost library is that it makes relatively complicated (yet powerful) template-based tools available to average developers. Which brings us to:
"This kind of programming relies too much on subtle complexities of the language definition...I wouldn't want to do that and wouldn't choose to use a library that may force me to do the same (which might happen if I decide to extend Boost's implementation)."
With every new language or paradigm, you'll find a group of stubborn people who refuse to adopt new techniques because it's "too much" of something for their tastes. For years, C programmers refused to adopt C++, because the code was "too abstracted" from the hardware. The same was said of C, by the assembly gurus. Now, you're giving iteration N+1 ("I'll never use it because the code syntax is too complicated"). It's just another form of Luddism.
"My rule of thumb as a developer: whatever you do, make sure that the next guy who wants to work with it doesn't have to get a Ph.D. in the subject in order to continue your work. In other words, keep your implementation simple and as mainstream as possible, so that it can easily be picked up by anyone."
Well then, I hope you're writing your code in BASIC, because there isn't another language in existence that is "simple" and "mainstream" enough to be "easily picked up by anyone".
And before you accuse me of being too literal, let me just pre-empt you: I get your point. You think C++ template code is too complicated. Fair enough. But you should realize that you're drawing a relatively arbitrary line in the sand, and your conception of "complicated" is likely to be eclipsed by whatever the next generation of programmers thinks is easy. And I can assure you that there are plenty of smart young programmers right now who are proficient and productive with tools like template metaprogramming and the boost library.
I'm not an idealist; there are plenty of things to complain about when it comes to C++, and any experienced C++ developer can rattle them off for an educated debate. The point is, dismissive rants by developers who simply think that the error messages of today's compilers are "too hard" are neither "insightful" nor "interesting" -- they're just whiny.
Ah, yes...the flamebait comment from the "expert" who doesn't understand how to use a powerful language feature. It never dies....
"(1) It's impossible to debug. You can't read the code. The debugger can't unravel the templated variables and stuff in any meaningful way for you. You can't even step through code, that's doing a supposedly simple operation like memory allocation!"
Well, rather than getting into a childish "can too!" debate with you, I'll posit that it must have been pretty damn hard to develop boost, if all of the things you mention were actually "impossible".
Better explanation: you don't know how to debug template code. Sad for you, but not a strike against the language.
"(2) Some compilers will choke on the code, or compile it wrong in subtle ways due to differing interpretations of some obscure section of the enormous C++ language spec."
Oooh...scary! Guess what, mate: some FORTRAN compilers choke on legitmate F77 code for the same reasons. This is nothing but a FUD tactic, dressed up as informed criticism.
Boost is well-tested against the major C++ compilers, and you can always find out the status of the various sub-libraries by going here.
"(3) The error messages from the compiler are useless. You have to run them through a filter to even figure out what they mean."
See #1, above.
"(3) Bugs in the library are very difficult to fix. Template metaprograms are essentially programs written in a functional language, except one that has horrible syntax. This is not the stuff that normal C++ programs are made of."
I think your second, third point says volumes about your problems: if you don't know how to use it, then it isn't "normal".
Sometimes, you have to learn new things to be productive in a new programming style. I could spend a great deal of time proving to you that there are many thousands of expert C++ developers who know how to do all of the things that you claim to be "impossible", but it isn't worth my time. I'm just hoping that a few knowledgeable developers will see this, and remove the ridiculous "Insightful" moderation from your rant....
Also, if you're interested in this stuff, you might want to check out Ebbesmeyer's website and newsletter about beachcombing: http://beachcombersalert.org/
You'd probably be surprised how many of the "brightest minds" spend eight hours a day doing needless grunt work to accomodate the many peculiarities of Microsoft Office.
While I am unsurprised to see the obligatory, reactionary, anti-intellectual slashdot response, I'd just like to take this opportunity to point out your blatant logical fallacy.
Logical fallacies: yet another part of a balanced, high-quality edumacation.
Are you fabricating quotes? I do not see such a sentence anywhere in the paper.
Copied verbatim from the methods section. Page 629, left column, mid-page. Sentence begins: "Given that the number of SNPs...."
That sounds quite contradictory. If SNPs aren't distributed according to a Poisson process, then the scan statistic found those deviations. The model appeared to work quite well, as the diagnostics indicate.
OK, I'm becoming convinced that you don't understand the paper:
The "scan statistic" is a measurement of deviation from a background model (in this case, a poisson process). The purpose of the statistic is to look for clusters of disease-associated SNPs. The assumption of a poisson process was nothing more than a good estimate for the naive model of DNA replication errors. The poisson process is the null hypothesis -- "interesting" SNPs deviate from this expectation.
"I've noticed some other bioinformatics types like yourself who aren't too well educated in any particular topic, so they become aggressive and make misinformed criticisms. When I was choosing grad schools, I went to a bioinformatics conference at Stanford and met a lot of unemployed bioinformatics graduates, so I steered clear from that field. But ok, I guess you're in 'computational biology', the latest marketing term for the same thing."
Cute. Well, fella...I don't know what to tell you. I certainly understand your paper better than you appear to understand it (I'm guessing that you were an undergrad contributor, so this makes some sense). In any case, I wouldn't be too quick with the snide remarks -- there are quite a few statisticians doing absolutely irrelevant biological research.
Perhaps you'll have a better sense of humility if you make it to the far side of a graduate program. Good luck.
"Ok, I'm not sure what you're arguing. As you say, you are not an expert in statistics, so I wouldn't know how to begin refuting your "pedantic" claim. I'll say that the paper was taken seriously enough to be published."
Uh...yeah. Did you even bother to read your own manuscript? From the methods:
"Given that the number of SNPs...is more likely to exponentially distributed, rather than uniformly distributed on chromosomes, a Poisson approximation model was used...."
In other words, you didn't find that SNPs are distributed according to a Poisson process, you chose a Poisson process to develop a tool (because it's a reasonable first assumption -- which is exactly what I said), and then found that the tool worked well enough as a background model (remember: your scan statistic is looking for SNP clusters that deviate from the poisson model). So the answer to my original question is quite simple: you didn't evaluate alternative models at all (other than the Fisher window method, which hardly counts).
Now, wouldn't it have been easier to simply answer my question with a straightforward, polite response, rather than writing me off as someone who wouldn't understand your brilliant work? I said that I wasn't an expert in single nucleotide polymorphisms -- I didn't say that I don't understand sequence analysis or statistics.
(Oh...by the way? Getting an article published in a journal is no guarantee that it's actually worth reading.)
If "point replication errors occur randomly and without mechanistic bias", that would imply a uniform distribution."
You're being pedantic. DNA replication takes place one nucleotide at a time, with some enzymatic procession rate. In the naive model, the polymerase moves at a constant rate from nucleotides 1->N, with some small probability of a replication error, expressed as a rate per N nucleotides. So far as I know, that's a poisson process.
Again, if you're a mathematical pedant, I suppose you could call these basic assumptions a "mechanistic bias," but I assumed that you wanted a biologist to take you seriously. There are literally dozens of ways that higher-order mechanisms could interfere with the ideal (for example, there are multiple origins of replication on every chromosome), and I wanted to know what other models you considered.
"I had a paper published about how DNA SNPs seem to follow a Poisson process in their distribution."
Isn't that pretty much what we would expect as the null hypothesis? It seems like the deviation from poisson would be the interesting phenomenon in this case....
More specifically: if point replication errors occur randomly and without mechanistic bias (i.e. they're unrelated to chromatin structure, or some other higher-order biological process), it seems like a poisson model would be the simplest description. If the poission wasn't your null hypothesis, what was?
Note: I am not an expert in this field, so it's a serious question.
One more thing: 6 miles a day, by bicycle, isn't very much exercise. I can do that in under a half an hour, and I'm not a very good bicyclist. If you want to lose weight, you need to to at least 30-45 minutes of strenuous aerobic exercise per session.
This may be why you never achieved significant weight loss through bicycling.
"I'm not saying exercise should not accompany diet, but, at least in the U.S., diet is almost always the big tamale. If you're eating 3500 calories a day (not at all hard to do), increasing your activity isn't going to do a thing."
1) You've completely missed my point: if forced to choose, increased exercise is better than dieting. Your body will adopt a starvation response to decreased caloric intake. Most people lose weight rapidly on a new diet (with significant water and/or muscle loss), then find the weight loss plateaus after a few weeks. This happens regardless of initial weight or metabolism -- it is the starvation response.
2) Your example is meaningless without qualifications. If you're a 200 lb man who exercises 3 times a week, 3500 calories is only a few hundred above baseline. One extra hour of moderate exercise per week would burn the difference.
"Case study: I have biked 6 miles a day for a couple years, but my weight has always been headed upward. Hmm. Ok, so I reduced my calorie intake to 1500/day, and in six months I'd lost 50 pounds. That was almost a year ago, and I've been at my ideal weight ever since."
Unless you're a woman, I can say that you're probably not eating enough. The American College of Sports Medicine recommends a minimum caloric intake of 1800 calories/day for men. If you're serious about your current diet, you should see a doctor or a nutritionist to have some basic tests done. You may very well be doing damage to your body.
"We all generally EAT TOO MUCH! Our guts are too big and so our hunger satisfaction signal is delayed. Working out and being active is good and all, but that's not the biggest part of what's wrong. It's WHAT we eat and how much of it we eat."
Actually, any doctor, physiologist or nutritionist will tell you that the problem has two parts: we don't exercise enough, and we eat too much. Both problems are equally important, and it's actually a far better idea to increase your activity than to drastically cut your caloric intake (if you're forced to choose). It's best to do both.
If you live a sedentary lifestyle but drastically cut calories, your body will eventually "decide" that you are starving, and will slow your metabolic rate to compensate (amongst other changes, such as the increase in serum cortisol levels, and the activation of lipid storage enzymes -- which essentially means that you'll begin to destroy muscle, in favor of preserving fat). This is why conventional diets do not work -- most people simply lose muscle mass (and/or water weight), eventually tire of starving themselves, and baloon back up to their pre-diet weight, with a lower lean body mass as a reward.
So, while the Big Mac culture is certainly a problem in the US, the only way to battle obesity in the long term is to encourage exercise. Dietary changes alone will not work.
Moderators, for the love of....don't just give high ratings to people who post technical-sounding gibberish!!
The parent post is spreading misinformation with regard to the link between libido and di-hydro-testosterone (you'll note that he got the abbreviation incorrect, and attached a spurious "5a" to the front, because he confused the name of the enzyme -- 5-alpha-reductase -- responsible for DHT formation with the chemical itself!)
I'm not going to claim that DHT isn't involved with male sex drive, but it's certainly not the "primary mechanism" behind male libido. One needs only refer to the volumes of studies done on the relationship between selective serotonin reuptake inhibior antidepressants ("SSRIs", e.g. Celexa, Prozac, Paxil) and libido suppression to see that the issue is more complicated than a single hormone imbalance.
What makes this really galling, however, is that the guy has the guts to criticize real scientific research before spouting this crap, and you folks take it as some sort of authority!
"The expected fuel economy gain is 30% over today's figures on the gasoline/FlexFuel-only AFM variant, approaching 30mpg for city driving. That's a damned good improvement. And when used with FlexFuel, they're using less fossil fuels - even including the fully burdened fossil fuel costs of ethanol - than Prius and Civic hybrid drivers, in addition to contributing to lower overall greenhouse gas emissions."
Uh, yeah....until Honda introduces an E85-capable hybrid. Then, SUVs will continue be the least fuel-efficient vehicles on the market.
No matter how you look at this, GM is shining a turd.
"Whether or not one likes or dislikes SUVs, or thinks people should be able to be told what types of vehicles they should or shouldn't be driving, or think subjective judgments can be simplistically made about what other people "need" or don't need, it's still an excellent step forward."
I'll grant that this is an important technological step forward, but I don't grant the greater implication: most people don't need to drive trucks. And no matter how many technoogical improvements are made to make light trucks more fuel-efficent, they'll still be less efficient than a smaller, lighter automobile with the same technology. It isn't a matter of "subjective judgment" -- it's a matter of physics.
(And not incidentally: we don't need to "tell" people what they "need" to drive. We can tax them based on the size and/or fuel-efficiency of their vehicle, and, like true conservatives, we'll "let the market work.")
"CPUs put a lot of stock in branch prediction; due to the nature of OOP languages like C++, Objective-C (I like this one), and D, this doesn't work. The way virtuals and class inheritence works, functions are necessarily dealt with as pointers; the function is pointed to by pointing to a master class object, basically....You can only justify OOP languages by saying that either A) the majority of the work doesn't involve making calls to other class members, and thus won't be hurt by this; or B) CPU speed doesn't matter. I hate argument (B); (A) I can accept, barely enough to tip my hat to you for having good software engineering sense."
Sorry, but could you repeat that? I can't hear you over the sound of your axe grinding.
Perhaps you don't know it, but you've (rather wordily) raised one of the classic red-herring arguments against C++: "Virtual Functions Are Slow." Now, sside from the fact that C++ doesn't force you to use virtual methods (which makes me wonder if you know anything about C++ at all), "good software engineering sense" will tell you that the 80/20 rule applies -- 80% of your execution time will be spent in 20% of your code.
In other words, if you're writing in a multi-paradigm language (like C++) you can avoid the cost of virtual function dispatch in the 20% of your code that is performance-critical, and for the rest, you can use virtual functions where they provide enhanced clarity and readability for your code. That's option (C), in case you were wondering.
So far, both of the high-rated responses have been correct -- you can either contact a commercial photo shop, and have the negatives scanned on a drum scanner (a $10,000+ piece of machinery, not incidentally), or you can do it yourself on a halfway-decent flatbed scanner.
Consider that even a low-end flatbed scanner these days can achieve 2000 dpi symmetric, which would allow you to print a 4x5 negative up to 26x33 inches at reasonably high quality (300dpi, which is enough for most uses). So the only question is, how big do you need to go?
There are other considerations than dpi, of course, but even a mid-range ($500-$800) scanner will have decent optics and color correction, and will be more than adequate for the job. Once you consider that most shops will charge you anywhere from $5-$40 per sheet to drum scan an negative, investing in a good flatbed scanner will quickly seem like a bargain....
Slashdot Mirror
I was at class when logins were launched...definitely registered the first day, though. It's a bit humbling to know that my login is almost old enough to drive.
Same here, Rob. You've had an amazing run. Good luck!
-Tim
Don't kid yourself -- there's a huge over-supply of graduate-level biologists with mathematical and computational training. The problem has nothing to do with the "softness" of the discipline, but rather, with the fact that academia pumps out doctorates at rates that can't be supported by industry.
Moreover, the button-pushers of the biology world are actually in a sweet spot, with regard to supply and demand. It's generally quite easy to obtain lab tech work with a BS (or even an AA). With one of these jobs, you'll live comfortably, work normal hours, and though you'll likely never lead a project, you'll earn respect and authority over time. An MS will bump you to a slightly higher salary, but it's questionable whether the gain is worth the opportunity cost (lost income, mainly).
By contrast, a PhD will leave you largely unemployable. You won't even be considered for most "PhD-level" positions, until you've completed an additional 2-4 years of post-doctoral "training", on top of the 5-7 years it takes to get the degree. And ironically, you won't even be able to get the lab tech work that you could have found with a BS/MS, because it is perceived as a "waste of talent" to put a PhD in a tech position, and most companies are fearful that you'll leave for greener pastures at the first opportunity (to be fair, this is probably true).
The OP is correct: if you're intelligent, and you're concerned about income, the biological sciences are a terrible place to be. Computer science is much more lucrative -- as are law, medicine, business and engineering. The perception that bio-tech is a job generator is largely a function of industry propaganda, and does not stand up to even casual scrutiny.
If you doubt me, go to the Science Magazine Careers Forum, and check out the number of truly sad stories in the field....
Have you ever actually *ridden* on a train in Japan? Outside of the major cities, it's pretty rural -- far more rural than anything but the most desolate parts of the American west. And ironically enough, it's actually *easier* to get higher speeds in rural areas, because there are fewer people to annoy, intersections to block, and so on.
Population density has nothing to do with the lack of efficient rail transit in the United States. We simply don't invest in it, because we're too busy subsidizing the air, car and trucking industries.
Well said, AC. I am perpetually annoyed by the number of people who bash on C++ simply because they don't want to take the time to learn a language that is more complicated than the ones they already know. There is more to the programming universe than Java and Perl!
Some goon gets a +5 moderation for admitting that the doesn't know C++ well enough to understand the compile errors he obtains, but the posts calling him out on his ignorance are met with "Troll" moderation. Welcome to slashdot.
"Complex syntax hinders communication between developers and increases defects. If everything else is equal (and it usually is), simpler syntax is better."
Granted, but the question remains: complex to whom?
I write template code (and use boost) on a regular basis, and I am familiar with the syntax. Once I got over the initial learning curve, I didn't find C++ template metaprogramming syntax to be particularly bizarre or difficult. I'm not trying to brag; I'm just pointing out that there is nothing inherently wrong with template code. It's understandable (and incredibly productive) if you take the time to learn to use the tool.
Again, this entire debate boils down to the question of what you find familiar and comfortable. If it were impossible for a programmer of normal intelligence to learn how to use boost, you might have a point; that's clearly not the case. In my experience, most complaints about C++ tend to be made by programmers who believe that their understanding of ALGOL-derived syntaxes should carry them comfortably through their programming careers. These types of people tend to get cranky and dismissive when they can't pick up a language in an afternoon.
"Fortunately, the next generation of programmers has many languages available that are not as complicated as C++. Progress in programming language design is indicated by syntactical simplicity, not complexity."
Demonstrably false. Progress in programming language (if it has been defined by a single metric at all), has been defined by increases in expressiveness, not simplicity. Cases in point: Java, Perl, Python and C are all significantly more syntactically complex (and expressive) than FORTRAN 77 and PASCAL.
"The GP sounds like he understands template metaprogramming very well."
Really? He was complaining about being unable to read the error messages.
"Nothing about his post suggests that he either a) says that he is an "expert" (as you are suggesting he does), or b) *isn't* enough of an expert to have formed a valid opinion on the subject."
You can't have it both ways: either the guy is informed enough to have an opinion on the matter, or he isn't. He made a post claiming that template metaprogramming is not "normal", while simultaneously making it absolutely clear that he didn't understand enough of what he was talking about to be even remotely qualified to advance a criticism.
Like it or not, there are quite a few C++ programmers out there who know how to work with complex templates. You may not know how, the GP poster may not know how, but that doesn't make the boost library "bad".
"It's only because of lots and lots of dedicated effort by Boost developers that Boost's very-complex template metaprogramming techniques work on a wide variety of popular compilers."
Straw-man. Any good library represents "lots and lots of dedicated effort" by developers, and almost by definition, wouldn't be good without that effort. Yet I doubt you're willing to abandon every domain-specific library, simply because you don't understand the code. Whether you realize it or not, your whole argument rests on that premise.
Keep in mind that the point of the Boost library is that it makes relatively complicated (yet powerful) template-based tools available to average developers. Which brings us to:
"This kind of programming relies too much on subtle complexities of the language definition...I wouldn't want to do that and wouldn't choose to use a library that may force me to do the same (which might happen if I decide to extend Boost's implementation)."
With every new language or paradigm, you'll find a group of stubborn people who refuse to adopt new techniques because it's "too much" of something for their tastes. For years, C programmers refused to adopt C++, because the code was "too abstracted" from the hardware. The same was said of C, by the assembly gurus. Now, you're giving iteration N+1 ("I'll never use it because the code syntax is too complicated"). It's just another form of Luddism.
"My rule of thumb as a developer: whatever you do, make sure that the next guy who wants to work with it doesn't have to get a Ph.D. in the subject in order to continue your work. In other words, keep your implementation simple and as mainstream as possible, so that it can easily be picked up by anyone."
Well then, I hope you're writing your code in BASIC, because there isn't another language in existence that is "simple" and "mainstream" enough to be "easily picked up by anyone".
And before you accuse me of being too literal, let me just pre-empt you: I get your point. You think C++ template code is too complicated. Fair enough. But you should realize that you're drawing a relatively arbitrary line in the sand, and your conception of "complicated" is likely to be eclipsed by whatever the next generation of programmers thinks is easy. And I can assure you that there are plenty of smart young programmers right now who are proficient and productive with tools like template metaprogramming and the boost library.
I'm not an idealist; there are plenty of things to complain about when it comes to C++, and any experienced C++ developer can rattle them off for an educated debate. The point is, dismissive rants by developers who simply think that the error messages of today's compilers are "too hard" are neither "insightful" nor "interesting" -- they're just whiny.
Ah, yes...the flamebait comment from the "expert" who doesn't understand how to use a powerful language feature. It never dies....
"(1) It's impossible to debug. You can't read the code. The debugger can't unravel the templated variables and stuff in any meaningful way for you. You can't even step through code, that's doing a supposedly simple operation like memory allocation!"
Well, rather than getting into a childish "can too!" debate with you, I'll posit that it must have been pretty damn hard to develop boost, if all of the things you mention were actually "impossible".
Better explanation: you don't know how to debug template code. Sad for you, but not a strike against the language.
"(2) Some compilers will choke on the code, or compile it wrong in subtle ways due to differing interpretations of some obscure section of the enormous C++ language spec."
Oooh...scary! Guess what, mate: some FORTRAN compilers choke on legitmate F77 code for the same reasons. This is nothing but a FUD tactic, dressed up as informed criticism.
Boost is well-tested against the major C++ compilers, and you can always find out the status of the various sub-libraries by going here.
"(3) The error messages from the compiler are useless. You have to run them through a filter to even figure out what they mean."
See #1, above.
"(3) Bugs in the library are very difficult to fix. Template metaprograms are essentially programs written in a functional language, except one that has horrible syntax. This is not the stuff that normal C++ programs are made of."
I think your second, third point says volumes about your problems: if you don't know how to use it, then it isn't "normal".
Sometimes, you have to learn new things to be productive in a new programming style. I could spend a great deal of time proving to you that there are many thousands of expert C++ developers who know how to do all of the things that you claim to be "impossible", but it isn't worth my time. I'm just hoping that a few knowledgeable developers will see this, and remove the ridiculous "Insightful" moderation from your rant....
Harper's did a long article on these in the January 2007 issue. If you're a subscriber, you can go to http://harpers.org/archive/2007/01/0081345 to read it.
Also, if you're interested in this stuff, you might want to check out Ebbesmeyer's website and newsletter about beachcombing: http://beachcombersalert.org/
You'd probably be surprised how many of the "brightest minds" spend eight hours a day doing needless grunt work to accomodate the many peculiarities of Microsoft Office.
While I am unsurprised to see the obligatory, reactionary, anti-intellectual slashdot response, I'd just like to take this opportunity to point out your blatant logical fallacy.
Logical fallacies: yet another part of a balanced, high-quality edumacation.
Yeah. Because everyone knows that anything worth doing can be done without edumacation.
Damned "reality-based" intellectuals....
Are you fabricating quotes? I do not see such a sentence anywhere in the paper.
Copied verbatim from the methods section. Page 629, left column, mid-page.
Sentence begins: "Given that the number of SNPs...."
That sounds quite contradictory. If SNPs aren't distributed according to a Poisson process, then the scan statistic found those deviations. The model appeared to work quite well, as the diagnostics indicate.
OK, I'm becoming convinced that you don't understand the paper:
The "scan statistic" is a measurement of deviation from a background model (in this case, a poisson process). The purpose of the statistic is to look for clusters of disease-associated SNPs. The assumption of a poisson process was nothing more than a good estimate for the naive model of DNA replication errors. The poisson process is the null hypothesis -- "interesting" SNPs deviate from this expectation.
"I've noticed some other bioinformatics types like yourself who aren't too well educated in any particular topic, so they become aggressive and make misinformed criticisms. When I was choosing grad schools, I went to a bioinformatics conference at Stanford and met a lot of unemployed bioinformatics graduates, so I steered clear from that field. But ok, I guess you're in 'computational biology', the latest marketing term for the same thing."
Cute. Well, fella...I don't know what to tell you. I certainly understand your paper better than you appear to understand it (I'm guessing that you were an undergrad contributor, so this makes some sense). In any case, I wouldn't be too quick with the snide remarks -- there are quite a few statisticians doing absolutely irrelevant biological research.
Perhaps you'll have a better sense of humility if you make it to the far side of a graduate program. Good luck.
"Ok, I'm not sure what you're arguing. As you say, you are not an expert in statistics, so I wouldn't know how to begin refuting your "pedantic" claim. I'll say that the paper was taken seriously enough to be published."
Uh...yeah. Did you even bother to read your own manuscript? From the methods:
"Given that the number of SNPs...is more likely to exponentially distributed, rather than uniformly distributed on chromosomes, a Poisson approximation model was used...."
In other words, you didn't find that SNPs are distributed according to a Poisson process, you chose a Poisson process to develop a tool (because it's a reasonable first assumption -- which is exactly what I said), and then found that the tool worked well enough as a background model (remember: your scan statistic is looking for SNP clusters that deviate from the poisson model). So the answer to my original question is quite simple: you didn't evaluate alternative models at all (other than the Fisher window method, which hardly counts).
Now, wouldn't it have been easier to simply answer my question with a straightforward, polite response, rather than writing me off as someone who wouldn't understand your brilliant work? I said that I wasn't an expert in single nucleotide polymorphisms -- I didn't say that I don't understand sequence analysis or statistics.
(Oh...by the way? Getting an article published in a journal is no guarantee that it's actually worth reading.)
"As you say, you are not an expert in statistics, so I wouldn't know how to begin refuting your "pedantic" claim."
Did I say that? Because that would be odd, considering that I'm very near the end of a PhD in computational biology.
Don't be a jerk. There are other smart people in the world who might actually understand your Special form Unique Genius.
If "point replication errors occur randomly and without mechanistic bias", that would imply a uniform distribution."
You're being pedantic. DNA replication takes place one nucleotide at a time, with some enzymatic procession rate. In the naive model, the polymerase moves at a constant rate from nucleotides 1->N, with some small probability of a replication error, expressed as a rate per N nucleotides. So far as I know, that's a poisson process.
Again, if you're a mathematical pedant, I suppose you could call these basic assumptions a "mechanistic bias," but I assumed that you wanted a biologist to take you seriously. There are literally dozens of ways that higher-order mechanisms could interfere with the ideal (for example, there are multiple origins of replication on every chromosome), and I wanted to know what other models you considered.
Would you like to answer the question now?
"I had a paper published about how DNA SNPs seem to follow a Poisson process in their distribution."
Isn't that pretty much what we would expect as the null hypothesis? It seems like the deviation from poisson would be the interesting phenomenon in this case....
More specifically: if point replication errors occur randomly and without mechanistic bias (i.e. they're unrelated to chromatin structure, or some other higher-order biological process), it seems like a poisson model would be the simplest description. If the poission wasn't your null hypothesis, what was?
Note: I am not an expert in this field, so it's a serious question.
One more thing: 6 miles a day, by bicycle, isn't very much exercise. I can do that in under a half an hour, and I'm not a very good bicyclist. If you want to lose weight, you need to to at least 30-45 minutes of strenuous aerobic exercise per session.
This may be why you never achieved significant weight loss through bicycling.
"I'm not saying exercise should not accompany diet, but, at least in the U.S., diet is almost always the big tamale. If you're eating 3500 calories a day (not at all hard to do), increasing your activity isn't going to do a thing."
1) You've completely missed my point: if forced to choose, increased exercise is better than dieting. Your body will adopt a starvation response to decreased caloric intake. Most people lose weight rapidly on a new diet (with significant water and/or muscle loss), then find the weight loss plateaus after a few weeks. This happens regardless of initial weight or metabolism -- it is the starvation response.
2) Your example is meaningless without qualifications. If you're a 200 lb man who exercises 3 times a week, 3500 calories is only a few hundred above baseline. One extra hour of moderate exercise per week would burn the difference.
"Case study: I have biked 6 miles a day for a couple years, but my weight has always been headed upward. Hmm. Ok, so I reduced my calorie intake to 1500/day, and in six months I'd lost 50 pounds. That was almost a year ago, and I've been at my ideal weight ever since."
Unless you're a woman, I can say that you're probably not eating enough. The American College of Sports Medicine recommends a minimum caloric intake of 1800 calories/day for men. If you're serious about your current diet, you should see a doctor or a nutritionist to have some basic tests done. You may very well be doing damage to your body.
"We all generally EAT TOO MUCH! Our guts are too big and so our hunger satisfaction signal is delayed. Working out and being active is good and all, but that's not the biggest part of what's wrong. It's WHAT we eat and how much of it we eat."
Actually, any doctor, physiologist or nutritionist will tell you that the problem has two parts: we don't exercise enough, and we eat too much. Both problems are equally important, and it's actually a far better idea to increase your activity than to drastically cut your caloric intake (if you're forced to choose). It's best to do both.
If you live a sedentary lifestyle but drastically cut calories, your body will eventually "decide" that you are starving, and will slow your metabolic rate to compensate (amongst other changes, such as the increase in serum cortisol levels, and the activation of lipid storage enzymes -- which essentially means that you'll begin to destroy muscle, in favor of preserving fat). This is why conventional diets do not work -- most people simply lose muscle mass (and/or water weight), eventually tire of starving themselves, and baloon back up to their pre-diet weight, with a lower lean body mass as a reward.
So, while the Big Mac culture is certainly a problem in the US, the only way to battle obesity in the long term is to encourage exercise. Dietary changes alone will not work.
Excuse the misplaced italics: di-hydro testosterone. The grandparent called it "de-hydro" testosterone, which is incorrect.
Moderators, for the love of....don't just give high ratings to people who post technical-sounding gibberish!!
The parent post is spreading misinformation with regard to the link between libido and di-hydro-testosterone (you'll note that he got the abbreviation incorrect, and attached a spurious "5a" to the front, because he confused the name of the enzyme -- 5-alpha-reductase -- responsible for DHT formation with the chemical itself!)
I'm not going to claim that DHT isn't involved with male sex drive, but it's certainly not the "primary mechanism" behind male libido. One needs only refer to the volumes of studies done on the relationship between selective serotonin reuptake inhibior antidepressants ("SSRIs", e.g. Celexa, Prozac, Paxil) and libido suppression to see that the issue is more complicated than a single hormone imbalance.
What makes this really galling, however, is that the guy has the guts to criticize real scientific research before spouting this crap, and you folks take it as some sort of authority!
"The expected fuel economy gain is 30% over today's figures on the gasoline/FlexFuel-only AFM variant, approaching 30mpg for city driving. That's a damned good improvement. And when used with FlexFuel, they're using less fossil fuels - even including the fully burdened fossil fuel costs of ethanol - than Prius and Civic hybrid drivers, in addition to contributing to lower overall greenhouse gas emissions."
Uh, yeah....until Honda introduces an E85-capable hybrid. Then, SUVs will continue be the least fuel-efficient vehicles on the market.
No matter how you look at this, GM is shining a turd.
"Whether or not one likes or dislikes SUVs, or thinks people should be able to be told what types of vehicles they should or shouldn't be driving, or think subjective judgments can be simplistically made about what other people "need" or don't need, it's still an excellent step forward."
I'll grant that this is an important technological step forward, but I don't grant the greater implication: most people don't need to drive trucks. And no matter how many technoogical improvements are made to make light trucks more fuel-efficent, they'll still be less efficient than a smaller, lighter automobile with the same technology. It isn't a matter of "subjective judgment" -- it's a matter of physics.
(And not incidentally: we don't need to "tell" people what they "need" to drive. We can tax them based on the size and/or fuel-efficiency of their vehicle, and, like true conservatives, we'll "let the market work.")
"CPUs put a lot of stock in branch prediction; due to the nature of OOP languages like C++, Objective-C (I like this one), and D, this doesn't work. The way virtuals and class inheritence works, functions are necessarily dealt with as pointers; the function is pointed to by pointing to a master class object, basically....You can only justify OOP languages by saying that either A) the majority of the work doesn't involve making calls to other class members, and thus won't be hurt by this; or B) CPU speed doesn't matter. I hate argument (B); (A) I can accept, barely enough to tip my hat to you for having good software engineering sense."
Sorry, but could you repeat that? I can't hear you over the sound of your axe grinding.
Perhaps you don't know it, but you've (rather wordily) raised one of the classic red-herring arguments against C++: "Virtual Functions Are Slow." Now, sside from the fact that C++ doesn't force you to use virtual methods (which makes me wonder if you know anything about C++ at all), "good software engineering sense" will tell you that the 80/20 rule applies -- 80% of your execution time will be spent in 20% of your code.
In other words, if you're writing in a multi-paradigm language (like C++) you can avoid the cost of virtual function dispatch in the 20% of your code that is performance-critical, and for the rest, you can use virtual functions where they provide enhanced clarity and readability for your code. That's option (C), in case you were wondering.
So far, both of the high-rated responses have been correct -- you can either contact a commercial photo shop, and have the negatives scanned on a drum scanner (a $10,000+ piece of machinery, not incidentally), or you can do it yourself on a halfway-decent flatbed scanner.
Consider that even a low-end flatbed scanner these days can achieve 2000 dpi symmetric, which would allow you to print a 4x5 negative up to 26x33 inches at reasonably high quality (300dpi, which is enough for most uses). So the only question is, how big do you need to go?
There are other considerations than dpi, of course, but even a mid-range ($500-$800) scanner will have decent optics and color correction, and will be more than adequate for the job. Once you consider that most shops will charge you anywhere from $5-$40 per sheet to drum scan an negative, investing in a good flatbed scanner will quickly seem like a bargain....