5 ms·

Stuff Goes Bad: Erlang in Anger

146 points by massung 12y ago

rdtsc 12y ago
This is the Missing Advanced Manual. Another great work by Fred Hebert, the author of http://learnyousomeerlang.com http://learnyousomeerlang.com . I like the exercises at end of chapters. Most of it is Erlang specific. But some chapters like "Planning For Overload" are interesting in general if you are running distributed system at scale.
ProfOak_ 12y ago
Looks like a really nice book, but I don't know a thing about Erlang.
- jacquesm 12y ago
  Then maybe it's time to expand your horizon a bit, he also has another book about Erlang that's a bit more beginner level: http://learnyousomeerlang.com/ http://learnyousomeerlang.com/ , I promise you that if you stick to it for a couple of weeks that you will not regret investing your time even if you never use Erlang professionally. Some of the philosophy behind it and the methodical way of attacking a problem will rub off and will make you a better programmer in the longer term.
jacquesm 12y ago
Erlang has the steepest learning curve of any programming language I've ever studied and Frank Hebert does an excellent job of grading that curve to the point where mere mortals can walk upwards. I'm tremendously grateful for all the work he's put into his books if you're going to spend time on Erlang start with 'Learn you some Erlang' and work your way into the libraries from there.
- therockhead 12y ago
  Do you mean the overall platform (OTP etc) ? While different, the Erlang language is small and much simpler than most programming languages IMHO.
  - jacquesm 12y ago
    OTP makes good sense to me because I built something along the same lines (in C, blasphemy) two years ago. Erlang the language has me tearing my hair out on occasion. That's not Erlangs fault, it's clearly mine, I grew up with 'BASIC' and moved on Assembler and from there to 'C' and a whole bunch of other imperative languages besides so functional languages are always harder for me to learn. That said I'm making good progress and Freds book was instrumental in that.
    - klibertp 12y ago
      Have you tried learning Prolog sometime? From experience: after a week or two with Prolog jumping back to Erlang was such a joy and relief that I picked it up very quickly :)
      - jacquesm 12y ago
        That sounds like stomping on your foot to make your headache go away.
        klibertp 12y ago
        Well, it's not that bad - the side-effect from learning Prolog is knowing a bit of Prolog, which is definitely much more useful than being hurt in the foot... No, seriously: Erlang shares some features with Prolog and learning Prolog lets you learn those features in isolation, without unnecessary distractions (like, say, processes and sending messages, or gen_server/gen_fsm/supervisors or rebar, applications, releases...). For example "=" not being an assignment, and pattern matching in general, starts making a lot of sense. Looping via tail-recursion (only), single assignment variables, and even some of Erlang syntax stops seeming intimidating quickly. I think the best metaphor for this would be learning to swim via jumping into a deep water, nearly drowning but learning the most important parts (like don't try to breathe underwater) and then going to shallow water to leisurely practice swimming there. It's certainly not for everyone, but for me it sped up the Erlang learning process by many weeks, if not months. (And also, I learned a bit of Prolog along the way.)
- kristoffer 12y ago
  I hate the misuse of the term "steep learning curve". If something has a steep learning curve it actually means you are learning it fast.
  - jacquesm 12y ago
    If you're a driver then you know that steep curves are taken slowly. (Unless you're going downhill, and in that case it's probably a sign that your brakes have failed...)
    - kristoffer 12y ago
      A learning curve is something specific. It is the level of knowledge over time. It has nothing to do with driving..
      - jacquesm 12y ago
        Ok, so let me make it graphical for you: If the 'Y' axis of the graph is expenditure of energy or time and the 'X' axis is progress in understanding then a steep curve indicates a lot of expenditure for relatively little progress. That's the 'informal use', as in, everyday communications such as, but not limited to comments in online communities.
        kristoffer 12y ago
        That is not a learning curve though... The "informal" usage stems from people not knowing what they are talking about. No need to perpetuate it if you know better.
      - grndn 12y ago
        Too late, I'm afraid. That horse has literally* bolted. The world has decided on what "steep" means now. * http://www.salon.com/2013/08/22/according_to_the_dictionary_literally_now_also_means_figuratively_newscred/ http://www.salon.com/2013/08/22/according_to_the_dictionary_...
    - 12y ago
- jrlocke 12y ago
  I believe you mean Fred Hebert.
  - jacquesm 12y ago
    Too late to edit, you're absolutely right.
tel 12y ago
This should be required reading for anyone building systems in which subcomponents can fail independently, Erlang or otherwise! The techniques are not always implementable in other languages, but the concepts behind them can and should be translated.

146 points by massung 12y ago

rdtsc 12y ago

This is the Missing Advanced Manual. Another great work by Fred Hebert, the author of http://learnyousomeerlang.com http://learnyousomeerlang.com . I like the exercises at end of chapters. Most of it is Erlang specific. But some chapters like "Planning For Overload" are interesting in general if you are running distributed system at scale.

ProfOak_ 12y ago

Looks like a really nice book, but I don't know a thing about Erlang.

jacquesm 12y ago
Then maybe it's time to expand your horizon a bit, he also has another book about Erlang that's a bit more beginner level: http://learnyousomeerlang.com/ http://learnyousomeerlang.com/ , I promise you that if you stick to it for a couple of weeks that you will not regret investing your time even if you never use Erlang professionally. Some of the philosophy behind it and the methodical way of attacking a problem will rub off and will make you a better programmer in the longer term.

jacquesm 12y ago

Erlang has the steepest learning curve of any programming language I've ever studied and Frank Hebert does an excellent job of grading that curve to the point where mere mortals can walk upwards. I'm tremendously grateful for all the work he's put into his books if you're going to spend time on Erlang start with 'Learn you some Erlang' and work your way into the libraries from there.

therockhead 12y ago
Do you mean the overall platform (OTP etc) ? While different, the Erlang language is small and much simpler than most programming languages IMHO.
- jacquesm 12y ago
  OTP makes good sense to me because I built something along the same lines (in C, blasphemy) two years ago. Erlang the language has me tearing my hair out on occasion. That's not Erlangs fault, it's clearly mine, I grew up with 'BASIC' and moved on Assembler and from there to 'C' and a whole bunch of other imperative languages besides so functional languages are always harder for me to learn. That said I'm making good progress and Freds book was instrumental in that.
  - klibertp 12y ago
    Have you tried learning Prolog sometime? From experience: after a week or two with Prolog jumping back to Erlang was such a joy and relief that I picked it up very quickly :)
    - jacquesm 12y ago
      That sounds like stomping on your foot to make your headache go away.
      - klibertp 12y ago
        Well, it's not that bad - the side-effect from learning Prolog is knowing a bit of Prolog, which is definitely much more useful than being hurt in the foot... No, seriously: Erlang shares some features with Prolog and learning Prolog lets you learn those features in isolation, without unnecessary distractions (like, say, processes and sending messages, or gen_server/gen_fsm/supervisors or rebar, applications, releases...). For example "=" not being an assignment, and pattern matching in general, starts making a lot of sense. Looping via tail-recursion (only), single assignment variables, and even some of Erlang syntax stops seeming intimidating quickly. I think the best metaphor for this would be learning to swim via jumping into a deep water, nearly drowning but learning the most important parts (like don't try to breathe underwater) and then going to shallow water to leisurely practice swimming there. It's certainly not for everyone, but for me it sped up the Erlang learning process by many weeks, if not months. (And also, I learned a bit of Prolog along the way.)
kristoffer 12y ago
I hate the misuse of the term "steep learning curve". If something has a steep learning curve it actually means you are learning it fast.
- jacquesm 12y ago
  If you're a driver then you know that steep curves are taken slowly. (Unless you're going downhill, and in that case it's probably a sign that your brakes have failed...)
  - kristoffer 12y ago
    A learning curve is something specific. It is the level of knowledge over time. It has nothing to do with driving..
    - jacquesm 12y ago
      Ok, so let me make it graphical for you: If the 'Y' axis of the graph is expenditure of energy or time and the 'X' axis is progress in understanding then a steep curve indicates a lot of expenditure for relatively little progress. That's the 'informal use', as in, everyday communications such as, but not limited to comments in online communities.
      - kristoffer 12y ago
        That is not a learning curve though... The "informal" usage stems from people not knowing what they are talking about. No need to perpetuate it if you know better.
    - grndn 12y ago
      Too late, I'm afraid. That horse has literally* bolted. The world has decided on what "steep" means now. * http://www.salon.com/2013/08/22/according_to_the_dictionary_literally_now_also_means_figuratively_newscred/ http://www.salon.com/2013/08/22/according_to_the_dictionary_...
  - 12y ago
jrlocke 12y ago
I believe you mean Fred Hebert.
- jacquesm 12y ago
  Too late to edit, you're absolutely right.

tel 12y ago

This should be required reading for anyone building systems in which subcomponents can fail independently, Erlang or otherwise! The techniques are not always implementable in other languages, but the concepts behind them can and should be translated.