Roffild's library - page 4
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Um... there are no columns at all in MQL. And MQL != SQL are completely different languages.
For those who know SQL, it's easier to sift out the best passes after optimization when their number exceeds 5000 passes. Implemented in TesterSql.mqh
Um... how could you think that...
The script for running Test Agents on amazon servers aws_ubuntu_user_data.sh now accounts for Rol and internal disks.
https://roffild.com/ru/
The script for running Test Agents on Amazon servers aws_ubuntu_user_data.sh now considers Role and internal disks.
Whose role? What do you mean by "internal disks"?
There is"Write data to MySQL format file", why not read from this format?
For the questions "what is Java, Spark, Alglib, Random Forest and Neural Networks? And how to live with it?" - Google
Alglib - single threaded
Spark - many computers.
The goal of the library is to use Spark (and in the future TensorFlow and MXNet) without MQL crutches
what tasks are solved with this product?
1) SQL costs money. Windows for SQL costs money. Amazon/Google servers cost money. Cost of one year with licenses? two months of google servers from his bonus - $300, that's 5-6 servers (on the bonus, the lifetime of servers on mining to ban from days to a couple of weeks). Ie per year, somewhere around $1.5-2k irrevocably "will fly away" in amazon/google.
2) A nifty laptop with a 6 core xeon and nvidia quadro $5k (add raid on m2 to basic configuration). MQL - free for end user. Embedded subd from hst files - for free (in OHLCV you can write anything you like, and access the data as a time series - i.e. zero intelligence costs). On the plus side, mobility. Computing power is sufficient.
3) Single-processor miniitx xeon + windows(license) + one/two powerful ATI graphics cards + monitors + oops. 5k$ . MQL - free for end user. Embedded subd from hst files - free (in OHLCV you can write everything you want, and access the data as a time series - i.e. zero intelligence costs). Pros - it is possible to make a configuration according to your own fenstyle. Minus, lack of mobility. Overdrive computing power.
Opencl speeds (video cards) are beyond competition in cost and processing power compared to "many computers".
The task at hand should repay irretrievable costs from 1), you need to know very well front/back java , sql and a bunch of other skills for big data straight out of your head - with such baggage of knowledge for a hired job with a salary in a bourgeois bank anywhere from 100k$ minimum, and google will probably give even more + access to computational resources.
Imho, to train on the cats and pump up skills for employment in a good position.
Roffild Library
I'm known to the MQL5 programming community as Roffild and this is my open-source library for MQL5. It's an attempt to implement features in MQL5 that became a standard for popular programming languages long ago. One idea is implemented in each file. The library is replenished as new features are needed.
Few people have tried to post the project on Github. There is no unified standard. MetaQuotes does not take into account the use of version control system when creating a project. For some reason, MetaQuotes programmers think that a project must be of one type. For small projects, which are published in CodeBase on MQL5.com, this division is reasonable. For medium and large projects, it is impossible to select one project type.
I even turned on the computer to reply, with a very limited internet traffic). It is not comfortable to write from a mobile phone.
With all due respect to Roffild's activities, and don't take that as a hitch, I read but don't understand. What for? [(c)A.Voznesensky.]
1. I've understood that a developer ported several libraries in MQL for himself and decided to make them available to the public. Everything is clear and normal. Why not. But there is a clear redundancy here - Roffild ports anything and everything.
2. A marketer would ask the question - target audience?
- The beginners and novices do not know MQL well enough, they have problems with adding the DLL. Do you think they would understand SQL or your documentation?
- Experienced MQL programmers may be able to use some of them. But only to a limited extent.
- Programmers in general (C++, etc.) - it is easier for them in the documentation to understand the source library and connect exactly what they need.
To sum up: in terms of marketing, we have several applications of several libraries from the whole set. Is that the reason for all the fuss?
3. All (or almost all) libraries are already ported to the same Python. Wouldn't it be easier to make just one port to a running Python application and use all the libraries at once, calmly and unhurriedly? Furthermore, all of the Python library ports are well tested and documented. In addition: Python is a scripting language and its time share in library execution is minimal. There is no impact on performance. Consider also that Python has full-fledged threads and many ports to MQL just hang without it, and develop when the train has already left...
I think in this case experienced programmers will understand you. And the proportion of applications of just one port to Python will be more than all your libraries in total.
Actually, there is still a "but" to your project, but that's already on occasion.
Regards.
1) SQL costs money. Windows for SQL costs money. Amazon/Google servers cost money. The cost of a year of usage with licenses? two months of google servers from its bonus is $300, that's 5-6 servers(on bonus the lifetime of servers on mining to ban from days to a couple of weeks). Ie per year, somewhere around $1.5-2k irrevocably "will fly away" in amazon/google.
2) A nifty laptop with 6-core xeon and nvidia quadro $5k (add raid on m2 to basic configuration). MQL - free for end user. Embedded subd from hst files - for free (in OHLCV you can write anything you like, and access the data as a time series - i.e. zero intelligence costs). On the plus side, mobility. Computing power is sufficient.
3) Single-processor miniitx xeon + windows(license) + one/two powerful ATI graphics cards + monitors + oops. 5k$ . MQL - free for end user. Embedded subd from hst files - free (in OHLCV you can write everything you want, and access the data as a time series - i.e. zero intelligence costs). Pros - it is possible to make a configuration according to your own fenstyle. Minus, lack of mobility. Overdrive computing power.
Opencl speeds (video cards) are beyond competition in cost and processing power compared to "many computers".
The task at hand should repay irretrievable costs from 1), you need to know very well front/back java , sql and a bunch of other skills for big data straight out of your head - with such baggage of knowledge for a hired job with a salary in a bourgeois bank anywhere from 100k$ minimum, and google will probably give even more + access to computational resources.
Imho, train on cats and pump up your skills to get a good job.
MySQL and SQLLite or similar. FREE. Servers may be rented at hourly rates. What's the point of renting for a whole year? It only takes me up to $200 a year.
OpenCL needs to be able to cook too. The long delay in sending data to the video card makes using OpenCL unprofitable.
When you figure it all out, only then can you really estimate the costs.
Yuriy Asaulenko:
3. All (or almost all) libraries are already ported to the same Python. Wouldn't it be easier to make just one port to a running Python application and use all the libraries at once in an easy and unhurried manner? Furthermore, all of the Python library ports are well tested and documented. In addition: Python is a scripting language and its time share in library execution is minimal. There is no impact on performance. Consider also that Python has full-fledged threads and many ports to MQL just hang without it, and develop when the train has already left...
Any third-party development is almost impossible to use in Test Agents. When trying to integrate with external developments you have to sacrifice speed, portability and optimisation. Why?
There's a joke among Linux users: "Put Linux, put Wine to run Kosynka".
And Python itself is a real slowpoke.
Python can be used to discover the patents of a model. When the model is found, it would be better to port it to MQL5 to test it in the Agents cloud.
My library does not depend on third-party DLL's, so it can be used in the code of any Expert Advisor, Indicator etc.
Any third-party development is almost impossible to use in Test Agents. When trying to integrate with external developments, you have to sacrifice speed, portability and optimisation. Why?
There's a joke among Linux users: "Put Linux, put Wine to run Kosynka".
And Python itself is a real slowpoke.
Python can be used to discover the patents of a model. When the model is found, it is better to port it to MQL5 to test it in the Agents cloud.
My library is independent of third-party DLL's, so it can be used in the code of any Expert Advisor, Indicator, etc.
The question of python speed and its comparison with C++, Sharp, etc. is not relevant at all. The question is whether its speed is sufficient for solving specific tasks. I believe that for the vast majority of tasks, even more than enough. Even for so-called scalping).
You believe that not using third party dlls is a boon. I believe this is more of a sectarianism, and, most importantly, a lot of unnecessary, and unnecessary work. All the libraries have already been made before us, and all that remains is to use them, with a minimum of time and effort. And this, by the way, is part of the modern programming paradigm. And, in general, it's not about Python. It's only as an example of an environment with lots of libraries.
The rejection of the DLL, the use of third-party software and other advances is justified solely for the sake of selling on the Market - such are the conditions.) There is simply no other rationale.
MySQL and SQLLite or similar. FREE OF CHARGE. Servers can be rented at an hourly rate. What's the point of renting for a whole year? It only takes me up to $200 a year.
OpenCL needs to be able to cook too. The long latency to send data to the video card makes it unprofitable to use OpenCL.
When you figure it all out, only then can you really estimate the costs.
- "Free" https://shop.oracle.com/apex/f?p=DSTORE:2:::NO:RIR,RP,2:PROD_HIER_ID:58095029061520477171389 means the cost of your personal knowledge(not wholesale from oracle) is three times higher . Why would you spend $200 to start a vegetable garden with an impliedly substantial amount of administrative knowledge???? If you have $200 to spend on google, then you can do the same thing in a couple of weekends on your beech/computer. The $300 example given is that a system with a local xeon is more productive than a bunch of vps. To solve a problem/project quickly, you need a couple of years of practice on similar tasks with dealing with unexpected bugs and bugs and redesigning from scratch. The entry threshold into your ideological scheme is far from free.
- Preparing OpenCL in MQL is less intellectually demanding than prof. level in xSQL(generally subdata). Collected code (not perfect) in codebase MQL and MQL development environment is sufficient and free (including servicedesk) for quick self-education, plus it shows immediately on screen the result without any fuss with windows, if you know the analogues (free development environment + example codes very much in one place) give. MQL ready-made cloud is cheaper than vps/many computers. So far, the price/performance options of MQL are beyond competition (and there is a place to spray venom over MQL).
So you're saying that transferring data to video card for OpenCL has higher latency than tcpip stack costs with network latency to vps????? (What substances do you use? ))))
- What exactly are you looking into? A cheap video card (e.g. not expensive ati radeon 580 8GB - $300) is more profitable than a vps, an old xeon is almost more profitable than a video card. They put graphics cards in laptops too. I can hardly imagine (or rather can't imagine) a computational task in MQL time series that would require more resources than a local MQL server/computer/cloud. Rare 64GB ecc + 2 xeon - $300 + PSU + monitor + keyboard mouse (another $100), the entire axis with guts in a frameisk, even use BASIC to write local data processing, it will be faster than vps+sql etc.