Hi,
On Jun 1, 2013, at 11:24 PM, Stéphane Ducasse stephane.ducasse@inria.fr wrote:
On May 31, 2013, at 7:09 AM, Tudor Girba tudor@tudorgirba.com wrote:
Hi,
First, thanks for having this discussion on the mailing list. Please keep it like that :).
I looked briefly at the code and here are some remarks:
- The PJASTNodeVisitor has confusing methods names (like they are also confusing in the RB visitor).
not anymore :) In 30 the RB visitor is clean and nice. The AST was cleaned too :)
Great. [ ... ]
- Ideally, the Visitor methods should mirror the AST hierarchy. For example:
acceptTypeArguments: aTypeArguments self flag: 'TODO'
should be: visitTypeArguments: aTypeArguments ^ self visitArgumentsNode: aTypeArguments
Why that? Why do you double all the visit method with Nodes: ?
I meant to say that to code should be:
visitTypeArguments: aTypeArguments ^ self visitArguments: aTypeArguments
so that the specific visit method calls by default the more generic one by following the AST hierarchy.
Cheers, Doru
Like this, I can subclass the visitor and provide a default behavior in visitNode: and only override a couple of other visiting methods where needed.
Cheers, Doru
On May 30, 2013, at 11:14 AM, Yuriy Tymchuk yuriy.tymchuk@me.com wrote:
Hi,
so there are a few parts.
- Visitors
I suggest that abstract visitor should guide you how to visit this node. For example just do 'visitNodes: aPrimaryWithSelectorsNode codeList' so when someone will overwrite it he will know that he should pay attention to codeList data. Also comment will be lovely to have and this is a tough part as almost no classes in PJ have comments.
2). AST nodes The reason why we create an AST is that PetitParser's native parse result are arrays, and we want more structured data. For instance this.getCommand().getContext().getUser().getActiveProfile() is a method call. It has a selector "getActiveProfile" it has no parameters and it's invoked on this.getCommand().getContext().getUser(). I may be wrong about this, but on the other hand current AST is constructed like this. So maybe if you need some other kind of AST you should create it and make one more subclass of parser that will construct that.
Also I'll sent a copy to moose-dev because maybe people there will have better vision.
Cheers Uko
On 29 трав. 2013, at 19:33, Chris Cunningham cunningham.cb@gmail.com wrote:
Hi,
I'm not sure what exactly you want out of the visitor pattern. It seems like the current accept.. methods (that aren't either 'TO DO' or #subclassResponsibility, that is) just ask you to #visitNode: or #visitNodes: for the various instance variables in the AST classes. Am I missing something here? I must say that I haven't used the visitor pattern with parsed results before, and am not sure what you want and/or need out of this.
As for PrimaryWithSelectorsNode, after looking back at it, I can see that I have obscured what the 'primary' was from when it was parsed out. however, in looking at what could constitute a primary (versus the selectors part), I find myself confused with what answering a primary would be. I could by most anything - for a single item (boolean, literal, variable) to something very complex (such as a string of identifiers hanging off of an identifier with or without message sends on the end). Instead I attempted to simplify it into an array of the parts so that I could iterate over them as needed. sometimes I would like to find the first in the list, sometimes the last, and sometimes I'd like to search the whole list for a specific call (if present) in the middle.
so, take the code: this.getCommand().getContext().getUser().getActiveProfile() The current PJPrimaryWithSelectorsNode will have these in the codeList array: this getCommand() getContext() getUser() getActiveProfile() I did change the behaviour of the parsed #primaryWithselectors - previously it would have bundled this and getCommand() together into one unit identified as 'primary' and all of the other methods as an array of 'selectors'.
similarly, this code: new SearchUrl(Search.class).set(parm1, p1).set(parm2, params.get2(p2)).set(parm3, p3).getHref() will return the codeList array: new SearchUrl(Search.class) set(parm1, p1) set(parm2, params.get2(p2)) set(parm3, p3).getHref() Previously, it would identify the first item of the array as the primary, and the rest as the selectors. Which does make a lot of sense to me.
finally, this code: this.context.getUser().getActiveProfile().getProfileProperties() will return in the code array: this context getUser() getActiveProfile() getProfileProperties() Previously, primary would have consisted of the array this context getUser() while selectors would have been the array of getActiveProfile() getProfileProperties()
I should note that I'm not a Java coder myself and am not clear on how Java coders identify the parts of their code. I have tried to mostly follow what was previously there in the parser as it was clearer than what I'd likely come up with. However, I can't really see myself why the 'primary' was primary in the previous examples - is it clear and I should revert back, and have the visitor visit the restored primary and selectors? or should I have the visitor visit each part of the stacked structure that the Java coder has presented us? Which way would you prefer it - I'll modify it to fit your desires.
Thanks, Chris
On Tue, May 28, 2013 at 10:55 PM, Yuriy Tymchuk yuriy.tymchuk@me.com wrote: HI,
I've checked the changes. `acceptPrimaryWithSelectorsNode:` has only a flag so I didn't get how should I accept it and started to check out what the idea is. I still don't get it. It has some code list that is usually an array. Why do we need that node? What code entity does it represent? It looks like instead of building an AST from parsed arrays we are wrapping them in other classes.
Yuriy
On 28 трав. 2013, at 21:30, Chris Cunningham cunningham.cb@gmail.com wrote:
Hi.
I've added the missing method (as well as related missing #acceptVisitor: methods on most of the other nodes that I've added) in the latest change.
If you have any other questions or requests, please let me know.
Thanks, Chris Cunningham
On Tue, May 28, 2013 at 7:45 AM, Yuriy Tymchuk yuriy.tymchuk@me.com wrote: Thank you
Надіслано з iPhone
28 трав. 2013 о 17:18 Chris cunningham.cb@gmail.com написав(ла):
Thanks for letting me know. I'll fix that today.
Sent from my iPhone
On May 28, 2013, at 1:59 AM, Yuriy Tymchuk yuriy.tymchuk@me.com wrote:
> Hi, > > your changes to PetitJava break my builds on fast. The problem is that `acceptVisitor:` of PJPrimaryWithSelectorsNode is not implementing `acceptVisitor: aVisitor` method, and so PJASTNodeVisitor is not implementing some visiting method that can give a hint on what should I do in my visitor subclass. > > Thank you for your contributions. > Uko
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-- www.tudorgirba.com
"Every now and then stop and ask yourself if the war you're fighting is the right one."
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
Moose-dev mailing list Moose-dev@iam.unibe.ch https://www.iam.unibe.ch/mailman/listinfo/moose-dev
-- www.tudorgirba.com
"Some battles are better lost than fought."