Dear members do not waste your time ... Regards, Ton.
----- Original Message ----- From: Ted Byers To: [email protected] Sent: Sunday, January 10, 2010 4:15 AM Subject: Re: [amibroker] A computer science related question on the AFL Language This is getting irritating. You are expressing your views in a manner that is rather insulting. I have my doubts about your claim to such experience. If you truly had experience with such large companies, you would have expressed your views with greater tact. While I would had made slightly different decisions on some aspects of AFL, I can understand the rationale behind those choices I would have made differently and I certainly would not criticize the product or the developer for the choices made. AFL is a specialty language, and so I would not expect the same things from it that I expect from the programming languages I normally use. In some respects it is analogous to VB. VB was originally designed as a scripting language for MS Office. In that domain it works very well. But as a general purpose programming language it is deeply flawed. It does what it was designed to do very well, but for other domains it is a poor choice. AFL, too, does what it does well, but I would not use it to develop a streaming video server. This reminds me of an experience I had early on in my career. I had to develop a product for the agricultural consulting industry. My product had several competitors, and their products were quite impressive. They had the look and feel of something designed by engineers for engineers. They were also commercial failures because they ignored the characteristics of the target market. The clients and the users were very different. The clients were farmers with earned doctorates in some aspect of agribusiness, with plenty of experience using IT. The users were secondary school graduates who had never turned on a computer. They were bright people, and could strip down, repair, and reassemble their farm equipment without any problems, but they were intimidated by IT. I developed a product that had a very simple interface, using jargon the users would be familiar with, and building in a little AI. That product was ridiculed by the competition, who falsely believed that the product was as simple as the user interface. However, the product was a commercial success BECAUSE THE USERS UNDERSTOOD IT. My product did all the same things that the competition did, but it's interface was designed to be intuitive for the end users. The feedback we received from the clients was that the users refused to use the competitors' products because they did not understand them, but they loved using the product I developed because they found it so intuitive that they did not need the user's manual. If you had the experience you claim to have, you would know that a product's design must take the end users needs into account, and you would have made allowance for that in your assessment of AmiBroker. This is a product that appears to have been around for a long time, and so it must have found a good fit between what it offers and what its users need and understand. Now, as for your claims about python, I am well aware that there are some 'libraries' that have been developed for it using fortran and C/C++, and so it is possible to get reasonably fast code from it. I am not impressed by it, though, since I can make my C++ code much faster than my fortran and C code, and I have used fortran for twice as long as I have C++. Going to something that is only as fast as good quality fortran code is for me a step backward. However, if I were developing a product that needed something analogous to AFL, I would rebuke any programmer working for me who suggested I provide support for python in that product. Neither Amibroker nor the products I develop need a full fledged general purpose programming language, although there is no question that a specialty scripting language is often useful. It is sufficient that one can develop plugins for AmiBroker, so one can use a general purpose language like C++ if one needs to do so. This has nothing to do with the nature of the python language itself. Rather, it has to do with the availability of capable programmers who know it and the rather small likelihood of customers asking for it. It is hard enough to find capable programmers who know mainstream languages like fortran or C++, or more recent languages like perl. When I look at undergraduate and college curriculae, I see some coverage of basic object oriented concepts, but very little dealing with generic programming and none at all dealing with more advanced concepts like template metaprogramming. Until there is significant demand for support for python bindings for one of my products and until there is a sufficient and reliable supply of capable python programmers, I will not be adding support for python bindings to any of my products. I am sorry I am not especially impressed by what looks to be one of your favourite programming languages, but if you want to advocate its use, you have to address the business aspects of such a decision in addition to any technical advantages the language offers. There is no point in me developing a product in python, or any other language, if it is too hard to find a capable python programming to maintain and extend it. And there is no commercial sense in providing support for python bindings if there are insufficient numbers of customers who could use it. In any event, this is hardly the place for you to try to sell the merits of python as most here are not likely to even understand the programming issues you and I may see. And I'd rather not see my mailbox cluttered by this stuff. I would rather see useful information about using AFL as it is to develop, and experiment with/backtest, trading systems. Cheers, Ted On Sat, Jan 9, 2010 at 9:20 PM, Potato Soup <[email protected]> wrote: Nobody's venting. And if you don't care about it, why are you posting? ---------------------------------------------------------------------------- From: "J Paul Buffon" <[email protected]> Date: Sat, 9 Jan 2010 21:15:51 -0500 To: <[email protected]> Subject: RE: [amibroker] A computer science related question on the AFL Language Well, if you have such a knowledge, send an e-mail may be privately to TJ, but I do not care about your venting here. In addition, not much respect for somebody posting with a fake name. No much respect neither for your multi-billion dollars programs, who put you in the class of Goldman Sachs, with so poor systems they had to get a AIG guarantee the taxpayers had to pay, thanks to their political influence. JP From: [email protected] [mailto:[email protected]] On Behalf Of Potato Soup Sent: 01/09/2010 8:45 PM To: [email protected] Subject: Re: [amibroker] A computer science related question on the AFL Language A very simple question to answer. I use your product because it is one of the few that offer decent built in drawing capabilities (pixel level). How much money I or my employers have is largely irrelevant, all tools are considered for the job. I'm sorry that you don't want to take some constructive criticism from other people in the field. That is unfortunate for your customers. I, nor anyone else has suggested that you failed in your original endeavor, or that AB/AFL is unworthy of any recognition. If you re-read my original posts you'll see that I mention that it has some positive attributes. But if you expect everyone to blindly praise you and suppress constructive criticism or advice than you will be disappointed. The best software takes the best ideas from everyone and synthesizes them together. I still have hope that AB can someday be the best. Cheers. ---------------------------------------------------------------------------- From: Tomasz Janeczko <[email protected]> To: [email protected] Sent: Sat, January 9, 2010 8:37:17 PM Subject: Re: [amibroker] A computer science related question on the AFL Language Hello, If you are so great and working for all those multi-billion institutions I am wondering what are you doing here, using so cheap and "flawed" system as you wrote in your original post. Surely for that amount of money that you have in hand you could hire best programmers in the world and write your own super-trouper language and all that stuff. So please give me a break. I wrote AFL for myself back in 1995 because I need tool for my own purposes. I decided to offer it for others and some people liked it. That's whole story. If you do not like it then search elsewhere or write your own for your multi-billion institution. And no thanks I do not need your advice. Best regards, Tomasz Janeczko amibroker.com On 2010-01-10 02:24, Potato Soup wrote: Sorry for adding my "twisted" opinion, which only has experience building trading systems for multi-billion dollar institutions. And Python's math modules that are implemented in Fortran and C are not "slow" In fact there are Python math implementations that run on GPUs. Python as a language can be optimized to be as fast as anything. I'm more than happy to offer help if the AB author needs assistance adopting more modern and "twisted" technologies. My goal in offering the feedback was to help improve the product. Cheers. ---------------------------------------------------------------------------- From: Tomasz Janeczko <[email protected]> Date: Sun, 10 Jan 2010 02:17:31 +0100 To: <[email protected]> Subject: Re: [amibroker] A computer science related question on the AFL Language Hello, Precisely. AFL is designed to be able to express trading system rules / indicators in easy and short form and to be fast. That's all. It is intended to be easy to use and fast thanks to array processing that general-purpose languages lack. Just compare Traders Tips formulas published on S&C site for various platforms (some of them using so called "standard" laguages) and you will quickly find which formula is the shortest and easiest to understand. That speaks for itself and has more weight that sombody's twisted opinion. If one wants full blown C++ - one can use it - just write your code in C++, compile it as a plugin and that's it. Python/Lua and all that stuff are slow compared to AFL array processing. And as far as object-oriented programming is considered, most people are not comfortable with it (and belive it or not but most people are NOT programmers). I can tell that because I hear a lot of feedback from users. Best regards, Tomasz Janeczko amibroker.com On 2010-01-09 14:57, Prashanth wrote: Its not a question of whether it can be improved or not. Its a question of how user friendly it is. Most traders are not programmers and hence complex coding is out of question for vast majority. Its this group that appreciates the easiness of coding in AFL as compared to other languages which may hold more potential but can be much more difficult to learn. I believe TJ has simpliefied as much as possible and maybe during that simplification process, there were some sacrifices that were done. Unless one is a hard core programmer, I feel AFL more than meets every specification. For those who like to use more tools, ADK is always there to use and create outside of AB what they desire to achieve. Cheers Prashanth ----- Original Message ----- From: Potato Soup To: AmiBroker (Discussion List) Sent: Saturday, January 09, 2010 19:15 PM Subject: Re: [amibroker] A computer science related question on the AFL Language Why design one when Python is free, as is Lua, Squirrel and other easy scripting languages? I have built many trading systems for hedge funds and big banks. Never once considered building a language with it. Sounds like you think AFL can't be improved? From a language perspective AFL has some good ideas and concepts but uneven execution. A lot of things feel incrementally added, whether they were or not I don't know. -------------------------------------------------------------------------- From: "J Paul Buffon" <[email protected]> Date: Sat, 9 Jan 2010 08:34:37 -0500 To: <[email protected]> Subject: RE: [amibroker] A computer science related question on the AFL Language Well, you should design one! As far as I am concerned, elegant and great price/quality ratio, the best in the market in that bracket. Part of my tool box and achieved north of 70% last year with it. JP From: [email protected] [mailto:[email protected]] On Behalf Of Potato Soup Sent: 01/09/2010 8:24 AM To: AmiBroker (Discussion List) Subject: Re: [amibroker] A computer science related question on the AFL Language AFL is an imperative language primarily with a dose of vector processing features that enable the terseness that you talk about. I would not describe it as an OO language in any way, just because it has OO bindings or provides access to objects. If you can not write an object then the OO syntax introduced in the backtester is syntactic sugar at best. I also wouldn't say it has anywhere near the power of C or C++ just because it offers some syntax similarities. Those languages derive their power mainly from their ability to access memory directly, and at the OS' discretion this means writing directly to hardware memory maps. Of course C++ takes things much further. But AFL doesn't give you anywhere the expressive data structure creation abilities that a true imperative or OO language would. Personally I feel AFL is a deeply flawed language that mixes constructs from Basic and C at very superficial levels. It provides its power from underneath the hood, not at the true language level. I would pay a lot of money for AB with Python as its language, using NumPy as the fast math and numerical processing underpinning. This is not to say that AFL doesn't have elegant concepts or advantages. It is just not a well designed language from the ground up. -----Original Message----- From: "cascade3891" <[email protected]> Date: Sat, 09 Jan 2010 08:26:59 To: <[email protected]> Subject: [amibroker] A computer science related question on the AFL Language Hi Amibroker community, I have specific questions about the AFL language, regarding where it stands within the computer language spectrum(s) and what effect that has on speed/performance, agility and modularity as well as its accuracy for readability and unit testing purposes. I know that AFL is not an object oriented programming language for the main part (however it does have some OO features like COM), does this make AFL primarily a functional programming language, a bit like OCaml?? Are functional programming languages better for financial trading applications? Where there is a need for speed, and quality stable code? AFL seems a lot brief in terms of the amount of code that you have to write (terse) ... this makes it more attractive for reading over and checking the code, and for backtesting purposes. I notice also that with AFL you don't have to declare data types, again making it much more efficient. Is there a drawback to using an OO code for financial trading systems/applications? I quite like the speed and terseness of the AFL language actually, and also since it has many similarities to C. But would there be any limitations to not being able to define classes and objects? I'm not an experience programmer so sorry if I sound green. Anyone have any ideas on Tomasz' original design philosophy when he set out creating the AFL language? to me it seems like he wanted to keep the power and similarities to C/C++ given the similar syntactical structure, whether because he knows that language well, or because he wanted it to be able to have the same sort of power, but he also seems to have kept in mind the needs for performance and stability, terseness for backtesting/speed purposes, and maybe also b/c most traders need to pick up the language, hence trying to make AFL easier to grasp. ------------------------------------ **** IMPORTANT PLEASE READ **** This group is for the discussion between users only. This is *NOT* technical support channel. TO GET TECHNICAL SUPPORT send an e-mail directly to SUPPORT {at} amibroker.com TO SUBMIT SUGGESTIONS please use FEEDBACK CENTER at http://www.amibroker.com/feedback/ (submissions sent via other channels won't be considered) For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG: http://www.amibroker.com/devlog/ Yahoo! Groups Links -- R.E.(Ted) Byers, Ph.D.,Ed.D. [email protected] CTO Merchant Services Corp. 350 Harry Walker Parkway North, Suite 8 Newmarket, Ontario L3Y 8L3
