Re: [nant-dev] Property Scoping
From: Ian MacLean [EMAIL PROTECTED] Sent: Friday, September 10, 2004 1:12 AM If this is still true then I wonder how the parallel task in Ant works around it. Properties in Ant are always read-only, so there shouldn't be any issue unless two parallel tasks attempt to create the same new property - an unlikely scenario. Gary --- This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170 Project Admins to receive an Apple iPod Mini FREE for your judgement on who ports your project to Linux PPC the best. Sponsored by IBM. Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
Andy Sipe wrote: I'm in the process of implementing a parallel type task (similiar to the existing Ant task). I've pretty much got it working and was playing with it in our current build file. The first thing I found was that properties that are declared within a target are global in scope. Needless to say, this caused a lot of problems when targets began to run at the same time. I did a bit of snooping in the archives and found that at one point (http://www.mail-archive.com/[EMAIL PROTECTED]/msg03300.html) there was talk of having the ability to scope properties to a given target vs being global. Has this been implemented in any way? No - not yet. I think we had basically decided that there was no pressing need for scoped properties. Perhaps you've just found one. I was under the impression that Ant properties also have global scope but maybe this has changed recently. From Ant - the definitive guide: the other prominent property characteristic is that properties are always global in scope. If this is still true then I wonder how the parallel task in Ant works around it. Ian -- Ian MacLean, Developer, ActiveState, a division of Sophos http://www.ActiveState.com --- This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170 Project Admins to receive an Apple iPod Mini FREE for your judgement on who ports your project to Linux PPC the best. Sponsored by IBM. Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
RE: [nant-dev] Property Scoping
What about scoped properties? - Mitch Denny - [EMAIL PROTECTED] - http://www.monash.net - +61 (414) 610141 - -Original Message- From: Ian MacLean [mailto:[EMAIL PROTECTED] Sent: Friday, 23 January 2004 7:07 AM To: Jaroslaw Kowalski Cc: Mitch Denny; Scott Hernandez; [EMAIL PROTECTED] Subject: Re: [nant-dev] Property Scoping To be honest I'd like to see us release a 1.0 with the current feature set before implementing somthing like typed properties. it would be kinda nice to unify properties and type references which are both essentially different types of variables. However I do feel that this is peripheral to what the majority of users want to do with nant. ie its quite a big change which doesn't actually make that much of a user-visible difference in many cases. Not that its not worth investigating just that it might be sensible to set this out past a 1.0 release. Ian Jaroslaw Kowalski wrote: Yeah, I was considering the same thing. I also wondered whether this could mean that there could be a unified type system. Filesets, string properties etc. You mean storing a fileset inside a property? Interesting idea. Gert, Ian, Scott - what do you think about typed properties? Jarek --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
To be honest I'd like to see us release a 1.0 with the current feature set before implementing somthing like typed properties. Agree. Still I'd like to see some verbosity patch and fileset extensions in 1.0 somehow. Hope those will find its way into 1.0 even they are not in release plan. it would be kinda nice to unify properties and type references which are both essentially different types of variables. However I do feel that this is peripheral to what the majority of users want to do with nant. ie its quite a big change which doesn't actually make that much of a user-visible difference in many cases. Not that its not worth investigating just that it might be sensible to set this out past a 1.0 release. To yours 2nd mail: Sure this is clean in a mathematical sense but personally I'd rather see a more verbose layout that actually gives me an idea of what files the fileset will match just by looking at it. Given that it can be hard to determine what a given fileset will match right now I imagine it will be that much more difficult trying to work out the result after a difference operation has been performed. Agree. I have more and more feeling that just add includes refid=/ into fileset should be enough. It is not type-general but... Martin --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
Yeah, I was considering the same thing. I also wondered whether this could mean that there could be a unified type system. Filesets, string properties etc. You mean storing a fileset inside a property? Interesting idea. Oh - we already mention it before. Would be great! And also + operator in expressions could be overriden to fileset merge :) And functions with fileset argument could exists instead of string as refid to fileset etc etc etc... Still we should be compatible with current id= syntax so typedefs should define property of its type? Sum: +1 for it! btw: maybe we will find some nicer syntax for merging here? e.g. property type=fileset name=id1 value=${id1+id2}/ Gert, Ian, Scott - what do you think about typed properties? Jarek --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
Oh - we already mention it before. Would be great! And also + operator in expressions could be overriden to fileset merge :) And functions with fileset argument could exists instead of string as refid to fileset etc etc etc... The overload is a bad idea. Currently it has problems with strings (concatenation) vs anything else (addition) and adding another exception would be even more confusing. Still we should be compatible with current id= syntax so typedefs should define property of its type? That would be consistent. We'd have a single build state (PropertyDictionary) instead of PropertyDictionary for simple values + DataTypeBaseDictionary for complex values. Sum: +1 for it! btw: maybe we will find some nicer syntax for merging here? e.g. property type=fileset name=id1 value=${id1+id2}/ I would propose: property name=id1 type=fileset value=${fileset::union(fs1, fs2)} / property name=id1 type=fileset value=${fileset::intersection(fs1, fs2)} / property name=id1 type=fileset value=${fileset::difference(fs1, fs2)} / Jarek --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
To be honest I'd like to see us release a 1.0 with the current feature set before implementing somthing like typed properties. it would be kinda nice to unify properties and type references which are both essentially different types of variables. However I do feel that this is peripheral to what the majority of users want to do with nant. ie its quite a big change which doesn't actually make that much of a user-visible difference in many cases. Not that its not worth investigating just that it might be sensible to set this out past a 1.0 release. Ian Jaroslaw Kowalski wrote: Yeah, I was considering the same thing. I also wondered whether this could mean that there could be a unified type system. Filesets, string properties etc. You mean storing a fileset inside a property? Interesting idea. Gert, Ian, Scott - what do you think about typed properties? Jarek --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
Sure this is clean in a mathematical sense but personally I'd rather see a more verbose layout that actually gives me an idea of what files the fileset will match just by looking at it. Given that it can be hard to determine what a given fileset will match right now I imagine it will be that much more difficult trying to work out the result after a difference operation has been performed. Ian I would propose: property name=id1 type=fileset value=${fileset::union(fs1, fs2)} / property name=id1 type=fileset value=${fileset::intersection(fs1, fs2)} / property name=id1 type=fileset value=${fileset::difference(fs1, fs2)} / Jarek --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
- Original Message - From: Ian MacLean [EMAIL PROTECTED] To: Jaroslaw Kowalski [EMAIL PROTECTED] Cc: Mitch Denny [EMAIL PROTECTED]; Scott Hernandez [EMAIL PROTECTED]; [EMAIL PROTECTED] Sent: Thursday, January 22, 2004 9:07 PM Subject: Re: [nant-dev] Property Scoping To be honest I'd like to see us release a 1.0 with the current feature set before implementing somthing like typed properties. I agree completely on this subject. There are things that are much more urgent than this (have a look at the releaseplan) ... Gert --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
Re: [nant-dev] Property Scoping
Seems like the flow scope should be called local in C#/programming terms, and local would be private/container-only scoping. Having global be the default is a good call, but only in come case, as you have identified in if//foreach//etc. I'm sure we will be able to say more with a patch; so we can test things out. If the changes are small enough, and defaults don't change existing functionality, I'm happy to put them in the head since we are still a little while from testing/beta'n this version. I'd be inclined to separate the two patches, one for scoped properties (which are core changes) and one for the new task def stuff. - Original Message - From: Mitch Denny [EMAIL PROTECTED] Hi folks, OK, I've got a bit of a prototype working for property scoping which so far appears to be non-breaking to existing scripts. It works like this: property name=x value=y accessibility=Global / Global is actually the default. If I had this: property name=x value=y accessibility=Local / It would mean that the property is accessible to all things in the current scope where a scope is defined by the current target (project for root level tasks) or TaskContainer. So this would cause an error in the expression evaluator: if test=true property name=x value=y accessibility=Local / /if echo message=${x} / Because x is not accessible outside of the scope defined by the if task container. This works with my earlier taskdef work too! Interestingly the following won't work. if test=true property name=x value=y accessibility=Local / if test=true echo message=${x} / /if /if Because local is local to the current task container. I introduced a third accessibility level called Flow which allows this to work. Remember that the default is Global when you are using the property / task, so it won't break anything. The way it works is that I have lots of PropertyDictionary objects attached to a hierarchy of Scope objects. The scope is updated when ever a build/target/task container starts or finishes. I also modified quite a bit of the implementation of PropertyDictionary so that it now stores a Property object as its value although the external interface is unaffected (cross fingers I didn't break anything). Now that I have done this, and if there is enough interest I'd like to propose that we do something like has been done for expression evaluation, take a branch and do some exploritory work on this where this = taskdef / and property scoping. Can I get a +1 or -1? --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
RE: [nant-dev] Property Scoping
Hiya Jarek, Yeah I wasn't sure about the nested if approach either, its easy enough to change the behavior (I think). By default its Global anyway, and flow fits the bill. Maybe Local should behave like Flow, and have one called Private or something. Shrug :) Come to think of it, there are very few cases where you wouldn't want to use either Global or Flow instead of Local. Need to think about where we want Flow properties to fall out of scope and cross breed it with Local. - Mitch Denny - [EMAIL PROTECTED] - http://www.monash.net - +61 (414) 610141 - -Original Message- From: Jaroslaw Kowalski [mailto:[EMAIL PROTECTED] Sent: Thursday, 22 January 2004 6:37 AM To: Mitch Denny; [EMAIL PROTECTED] Subject: Re: [nant-dev] Property Scoping Hi Mitch! property name=x value=y accessibility=Global / Global is actually the default. If I had this: property name=x value=y accessibility=Local / It would mean that the property is accessible to all things in the current scope where a scope is defined by the current target (project for root level tasks) or TaskContainer. So this would cause an error in the expression evaluator: if test=true property name=x value=y accessibility=Local / /if echo message=${x} / That's intuitive. You usually expect variable to be inaccessible when it leaves the scope. Because x is not accessible outside of the scope defined by the if task container. This works with my earlier taskdef work too! Interestingly the following won't work. if test=true property name=x value=y accessibility=Local / if test=true echo message=${x} / /if /if This one is counter-intuitive. Most languages use local for what you've called flow. What is the use for local scoping anyway? Because local is local to the current task container. I introduced a third accessibility level called Flow which allows this to work. Remember that the default is Global when you are using the property / task, so it won't break anything. The way it works is that I have lots of PropertyDictionary objects attached to a hierarchy of Scope objects. The scope is updated when ever a build/target/task container starts or finishes. I haven't noticed a patch attached, but I don't know why do you want to store multiple dictionaries? Usually the this kind of processing can be done entirely on the evaluation stack. I also modified quite a bit of the implementation of PropertyDictionary so that it now stores a Property object as its value although the external interface is unaffected (cross fingers I didn't break anything). Now that I have done this, and if there is enough interest I'd like to propose that we do something like has been done for expression evaluation, take a branch and do some exploritory work on this where this = taskdef / and property scoping. This definitely needs some thought. +1 for the branch idea. Can I get a +1 or -1? +2/3 Jarek --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
RE: [nant-dev] Property Scoping
Scott, I'd really prefer to branch on this - it has the ability to really break things. One of the first changes is changing the PropertyDictionary to to store Property objects instead of string values. The property object is where the accessibility level is stored. While I am really keen to put this in, I realise that this may not be stable before the next release. One of the key drivers for the scoped properties (for me anyway) is the inline task definitions. That's why I wanted to see them together. Scoped properties would need to be implemented first mind. - Mitch Denny - [EMAIL PROTECTED] - http://www.monash.net - +61 (414) 610141 - -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Scott Hernandez Sent: Thursday, 22 January 2004 7:37 AM To: [EMAIL PROTECTED] Subject: Re: [nant-dev] Property Scoping Seems like the flow scope should be called local in C#/programming terms, and local would be private/container-only scoping. Having global be the default is a good call, but only in come case, as you have identified in if//foreach//etc. I'm sure we will be able to say more with a patch; so we can test things out. If the changes are small enough, and defaults don't change existing functionality, I'm happy to put them in the head since we are still a little while from testing/beta'n this version. I'd be inclined to separate the two patches, one for scoped properties (which are core changes) and one for the new task def stuff. - Original Message - From: Mitch Denny [EMAIL PROTECTED] Hi folks, OK, I've got a bit of a prototype working for property scoping which so far appears to be non-breaking to existing scripts. It works like this: property name=x value=y accessibility=Global / Global is actually the default. If I had this: property name=x value=y accessibility=Local / It would mean that the property is accessible to all things in the current scope where a scope is defined by the current target (project for root level tasks) or TaskContainer. So this would cause an error in the expression evaluator: if test=true property name=x value=y accessibility=Local / /if echo message=${x} / Because x is not accessible outside of the scope defined by the if task container. This works with my earlier taskdef work too! Interestingly the following won't work. if test=true property name=x value=y accessibility=Local / if test=true echo message=${x} / /if /if Because local is local to the current task container. I introduced a third accessibility level called Flow which allows this to work. Remember that the default is Global when you are using the property / task, so it won't break anything. The way it works is that I have lots of PropertyDictionary objects attached to a hierarchy of Scope objects. The scope is updated when ever a build/target/task container starts or finishes. I also modified quite a bit of the implementation of PropertyDictionary so that it now stores a Property object as its value although the external interface is unaffected (cross fingers I didn't break anything). Now that I have done this, and if there is enough interest I'd like to propose that we do something like has been done for expression evaluation, take a branch and do some exploritory work on this where this = taskdef / and property scoping. Can I get a +1 or -1? --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers
RE: [nant-dev] Property Scoping
Hiya, Yeah, I was considering the same thing. I also wondered whether this could mean that there could be a unified type system. Filesets, string properties etc. - Mitch Denny - [EMAIL PROTECTED] - http://www.monash.net - +61 (414) 610141 - -Original Message- From: Jaroslaw Kowalski [mailto:[EMAIL PROTECTED] Sent: Thursday, 22 January 2004 7:50 AM To: Mitch Denny; Scott Hernandez; [EMAIL PROTECTED] Subject: Re: [nant-dev] Property Scoping Scott, I'd really prefer to branch on this - it has the ability to really break things. One of the first changes is changing the PropertyDictionary to to store Property objects instead of string values. The property object is where the accessibility level is stored. The branch is a good idea. I'd like to evaluate another issue: can we have TYPED properties? Like this: property name=counter type=integer value=0 / This way you can write: if test=${counter + 1 = 100} /if instead of: if test=${convert::to-int(counter) + 1 = 100} /if Once a property has its type set, it cannot be re-typed. Every time you store a value in such a property, it is checked for type compatibility. Storing data type (optional - would default to string) in a PropertyDictionary would help here a lot. Jarek --- The SF.Net email is sponsored by EclipseCon 2004 Premiere Conference on Open Tools Development and Integration See the breadth of Eclipse activity. February 3-5 in Anaheim, CA. http://www.eclipsecon.org/osdn ___ nant-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/nant-developers