PowerLanguage Editor: Revision Control

Questions about MultiCharts and user contributed studies.
User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

PowerLanguage Editor: Revision Control

Postby Bruce DeVault » 25 Feb 2010

It would be helpful for the PowerLanguage editor to incorporate revision control e.g. track previous versions of a given function/study/signal. In this way, it would be possible to go back to a previous version.

In a related note to this, it may also be helpful to allow specifying a given version to run on the format dialog, so that edits can be made to a version in development and tested separately, without affecting production charts.

User avatar
TJ
Posts: 7284
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 992 times
Been thanked: 2071 times

Postby TJ » 25 Feb 2010

I think this is easily accomplished... with only a little programming

when the user press the "Compile" key,

PLEditor would simply rename the existing version to a new name.

(eg. append a number to the end of the indicator name, myindicator_1)




p.s. disclaimer: everything looks simple and easy from a distance. ;-)

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 25 Feb 2010

Welcome to a world in which there's a myindicator_1234 and your scrolling list of indicators when you go to add a study is thousands of items long, though.

Some kind of actual management of this would likely be required.

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 25 Feb 2010

One use of a simple revision control is to roll back to a prior version very quickly for the current trading day. It happened to me just this week. I used the replay to work on a study over the weekend and it was testing fine. I ran it on Monday and for some reason it was not (more work was needed to deal with situations where barstatus is not equal to 2 only). So I quickly rolled it back for the trading day on Monday and completed it later when I had a chance to give it some thought (ready to minimize the loss of live trading time to test it). With a revison control system I am suggesting I would just choose to have MC run off the Friday version instead of the one that compiled clean on Sunday. So the library would store not only the source could but the link as well. If not maybe it would just recompile direct from that older version. With the current setup I had to rename the current version and import the old version and make sure I kept it all straight. You get all sorts of temp names in the archives that way.

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 25 Feb 2010

I am thinking along the lines of the Editor would show at the top the version you are working on and the version in production. The MC Study list would show the number of versions in the library and the version running in production now. The export would export all versions and import all versions. The open would allow you to bring in an old version (2 back) and create a new version (maybe restricting it to new versions only). Maybe the user could grab all of a version 2 back then insert some of a version only 1 back (not sure how that would work).

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 25 Feb 2010

The simplest, common sense way this could work would be that:

(1) each version (each unique changed source code file that is saved) has its source code retained, organized by version for the indicator.
(2) the editor allows you to tag versions with a particular tag, such as "production" or "beta test", etc. and for any version that is tagged, the compiled object code would also be saved.
(3) one of the tags would be defined in the editor as the default for this indicator.
(4) in a chart, when you apply the indicator, it would default to the version that has the tag that is the default, but allow you to pick a different tag e.g. choose the "beta test" one instead.

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 25 Feb 2010

Yes I was thinking pretty much along that line although I took a break and had not gotten that far into it. I was thinking that the export would store the currently running version and that would be picked up on the inport to another machine. You could simply go in and change it. Also when you go in to open a study it would show you the total revisions and the one flagged as running on the chart(s) now. Maybe it could even be changed there. Maybe that would be a special popup with the list of revisions and you you change the one you want to run on. Making that change would automatically kick off the "loading" process in MC just as we currently get the "verifying" process right now. The list of revisions would of course show if the revision was compiled successfully yet or not. LOL, I can see TSS rolling their eyese right now :D, thinking "Hey guys, as if we don't already have enough work".

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 25 Feb 2010

I agree this is not and should not be a high priority for TS Support - no other trading platform has revision control, yet there are plenty of other trading platforms that have features that need to be accomodated here. It's simply the sort of thing that when they're redoing the PL Editor they should keep in mind because it may be something they can accomodate during such a revision (for instance to fix the handling for long files, etc.)

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 25 Feb 2010

100% Agreed. When the time comes that they do have the time they can search back for the ideas (or use whatever method they use).

Getting back to the point I made above about importing, currently an import flags that study as not verified and it is up to the user to verify it before opening the workspace. So to keep that concept, if three revisions existed in the inported study, it should remove the flag and it is up to the user to select the one they want to be running on this machine. Much like "compile all" they would have a an option to flag all the latest version for running on the machine and the user would then decide to go back and make a few specialized adjustments if they wanted.

I used two revision control systems with unix. One was RCS and I forget the name of the other. It started with S. SCCS I think. RCS could merge if I remember. It has been a while.

wegi
Posts: 121
Joined: 02 Jun 2009
Has thanked: 5 times
Been thanked: 12 times

Postby wegi » 25 Feb 2010

Hi,


i use subversion as my revision control system, in combination with tortoise, which is an windows
fileexplorer extension.
The only problem is, that it is not integrated in the PL-Editor, so i have to do this with the explorer.
But this saves me a lot of time.
You can:
- make new revision of files, and swith back to older revisions.
- make a "Tag". That means that you "mark" several revision of files
which should belong together: function_A version 1.3 and function_B 1.4
So you can swith back to older "Tags" and you will get all the corresponding files.
I do this, if i have a system which i am trading in praxis, then i mark all files,
which belong to this system with an tag: "praxis system date...".
And this saves me a lot of time of searching errors !!

So i don´t think that MC needs to code a new solution, instead it should allow to integrate any
revision control system, such as subversion, which is very common !

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 25 Feb 2010

Hi wegi

Do you cut and paste the code in to the revision control system with each change or does it pick up the PLA file directoy some how.

John.

wegi
Posts: 121
Joined: 02 Jun 2009
Has thanked: 5 times
Been thanked: 12 times

Postby wegi » 26 Feb 2010

You don´t need to copy & paste code.
The whole files are stored in the revision controll system: subversion

You need the tool tortoise.
This is an extension for the windows file explorer.
you navigate to the location of your PLE file, click rigth with the mouse on it and you will see the revision controll extensions: update, commit, browse ...

you can also click on the whole directory folder and "commit" all changes.
this will upload every changed file, you can provide an comment and a tag name. With this tag name, you can download the files, if needed.

tortoise includes:
- browsing the revision repository
- Tagging files
- compare two revision
... lot more, but this maybe are the most importand you will need.

if you install subversion, you simply have the revision controll system, which you can controll with "shell commands".
So, tortoise is an needed extension.

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 26 Feb 2010

I guess the question is that in Unix I could see the text in any simple text editor and the RCS stored it so it could compare code, merge it, etc. I have no idea where MC's EL code is before it is stored to a PLA archive nor what format it is in (ASCII?). Storing an archive in an external revision control system does not make much sense to me because first off I don't think it is stored in ASCII and secondly an archive is several studies of different names where as revision control deals with one study at a time for the purpose of helping to manage different versions.
Last edited by bowlesj3 on 26 Feb 2010, edited 2 times in total.

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 26 Feb 2010

It's stored in tokenized form. Binaries can be stored in version control - they just don't benefit so much from change analysis nor from the storage savings of just storing the changes.

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 26 Feb 2010

Thanks Bruce. I figured something like that. It was the reason I asked if wegi was using cut and paste.

wegi
Posts: 121
Joined: 02 Jun 2009
Has thanked: 5 times
Been thanked: 12 times

Postby wegi » 28 Feb 2010

My MC Files are stored at: all users\...\multicharts\study server\Studies\SrcEL\

There i find my strategies and signals, the files ends with .el, but they are text-files, which i upload to subversion.

but i have to say, that MC is not my main development software, i do all my studies, backtest ... with tradesignal. only the strategies a wanted to trade are coded than in MC.
So at the moment i ony upload my MC files, i was never in the situation to swith back to an older version. maybe i have to test this, because the way i do this is wrong :oops:

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 28 Feb 2010

but i have to say, that MC is not my main development software, i do all my studies, backtest ... with tradesignal. only the strategies a wanted to trade are coded than in MC.
Could you elaborate please on why you use TradeSignal for this? It might be instructive in terms of helping TS Support to take MC in a direction that leads to better feature completeness.

wegi
Posts: 121
Joined: 02 Jun 2009
Has thanked: 5 times
Been thanked: 12 times

Postby wegi » 28 Feb 2010

Could you elaborate please on why you use TradeSignal for this?
I will try, but the main reason is, that i am working with tradesignal several years now and i am very familiar with it.

What i am missing in MC:
- drawing from signals
- global variables
- access higher TFs without an second symbol in the chart.
- useability (see my Inputs at the right side, no need to open an separate window)

But the main big advantage is that in tradesignal i can work modular.
That means i put on 3 signals, lets say 3 stop-modules.
Than this modules are processed orderd, in the order they where put on the chart. In module 2 i can take the results from modul 1 and do something.

If i want do do this in MC i have to code everything in functions.
Lets say a trend indicator based on a simple moving average. I need an indicator to draw the information on the chart and code the logic in a function. Then i need a signal script and use the function. But i have the same inputs in the Signal and the Indicator and the code has to be executed twice.

I have several trend modules. In my signal script i have only three informations from this moduls: Trend long, Trend Short, trend flat.
It is much better to test trend or stop moduls and i don´t need all the inputs in my main strategy. Each module has an input: Active: yes/no.
so i can test with "drag and drop" my modules on the chart.

Buy the way, this is very usefull for portfolio trading !
A stock portfolio with 100 Stocks. The first modul on the portfolio calculates possible entries. the second modul now knows: 20 possible entries, but only 5 allowed per input, so only 5 order for stock 12345.

Not sure how the signals and indicators are processed in MC ?

wegi
Posts: 121
Joined: 02 Jun 2009
Has thanked: 5 times
Been thanked: 12 times

Postby wegi » 28 Feb 2010

I think with global variables and an ordered processing of signals and scipts, multicharts would do an great step !

Take a look at the forum on how many people ask for global variables to share informations and to syncronize scripts.

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 28 Feb 2010

See also as a minor suggestion: http://forum.tssupport.com/viewtopic.php?t=7191.

bowlesj3
Posts: 2125
Joined: 21 Jul 2007
Has thanked: 206 times
Been thanked: 420 times

Postby bowlesj3 » 28 Feb 2010

MC does have global variables (or can use them). Are you referring to public variables meaning that you can put variables in a module to be shared by all EL code in a chart or maybe in a workspace or maybe even in all of MC. Public variables would not be shared outside of MC as global variables can be. Public variables (if they existed in MC) would not need to be attached to the back end of an MC function to be used by that function. This is what you can do in VBA and I assume many other languages too. In cobol everything is public.

Oh by the way, thanks for pointing out where the EL source code is wegi

wegi
Posts: 121
Joined: 02 Jun 2009
Has thanked: 5 times
Been thanked: 12 times

Postby wegi » 28 Feb 2010

The public variables should be shared only in the chart between all "scripts".
And not between all charts in MC.

The current global variables are only available if you use a dll extension, e.g. pushpop.dll. But i can´t use them at the same bar, not for live trading, i don´t trust them, because MC don´t have any order, how the scripts are processed.

But this goes out of the main topic here, maybe we should start an new thread.

User avatar
TJ
Posts: 7284
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 992 times
Been thanked: 2071 times

Postby TJ » 28 Feb 2010

Could you elaborate please on why you use TradeSignal for this?
I will try, but the main reason is, that i am working with tradesignal several years now and i am very familiar with it.

What i am missing in MC:
- drawing from signals
...

you can draw from signals.

ie. ARW_NEW, TL_NEW, TEXT_NEW


(ps. you cannot plot in signal)

.
Last edited by TJ on 28 Feb 2010, edited 2 times in total.

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 28 Feb 2010

The inability to plot from signals is a major source of pain. It would be good to have TS Support simply address this directly.

wegi
Posts: 121
Joined: 02 Jun 2009
Has thanked: 5 times
Been thanked: 12 times

Postby wegi » 28 Feb 2010

@TJ, i mean plot, not draw.

I asked MC if i can DRAW about 1.000.000 Objects, i think you know the answere :-).
1.000.000 are exact 5 informations (drawings) per Bar if you have 200.000 Bars in a chart. And this is about 5min Data, 3 years.

But i read here in the forum, that plotting from signals will come.

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 04 Mar 2010

You would probably be better served in the short term to use ELC or ADE for this, so you could plot from a strategy on historical bars.

janus
Posts: 771
Joined: 25 May 2009
Has thanked: 45 times
Been thanked: 91 times

Postby janus » 04 Mar 2010

But i read here in the forum, that plotting from signals will come.
Now that would be great. It's tiring to have two studies, one to draw indicators and one to place orders based on the same indicator logic. Even if the computations are in a function that's used by both studies, the code is executed twice. What a waste.


Return to “MultiCharts”