[vwnc] parser for variable types in class comments

Rob Vens schrieb am 02.03.2008 11:00:

> For a tool I am writing to generate XMI exports from Smalltalk (currently
> for import into case tools, specifically Sparxsystems Enterprise Architect)


By the way, thank you very much for publishing it in the public
repository! It may help us in our current project.


Joachim Geidel

Rob,

 

You are trying to parse comments from XML, so this isn't directly relevant; however...

 

You could try a variation of the code in the attached file. The code is part of a scanner that efficiently compares code while ignoring formatting differences. I've only included the code relevant for extracting comments according to Smalltalk syntax. You might be able to customize for your parsing needs.

 

I'm surprised you need to parse the information you are looking for--instead of just generating it from code definitions.

 

Paul Baumann 

 

---

From: [hidden email] [mailto:[hidden email]] On Behalf Of Rob Vens
Sent: Sunday, March 02, 2008 5:00 AM
To: VWNC
Subject: [vwnc] parser for variable types in class comments

Has anybody done some work on parsing the class comments?
This uses a semi-formal notation, like:

Subclasses should implement:
accessing
   getPath

Instance Variables:
   name <String> the name of the thing
   employers <Set of: Person> a collection
   maritalStatus <'married' | 'single'>

etc.
For a tool I am writing to generate XMI exports from Smalltalk (currently for import into case tools, specifically Sparxsystems Enterprise Architect) I am using a brute force approach to parse this information, but possibly someone has done some work on this I am not aware of. I am not using the ParserCompiler toolkit or SmaCC for now.
The export to UML case tools is especially handy for eXploratory Modeling sessions, where Smalltalk is used to model the customers problem domain.

 

---

This message may contain confidential information and is intended for specific recipients unless explicitly noted otherwise. If you have reason to believe you are not an intended recipient of this message, please delete it and notify the sender. This message may not represent the opinion of IntercontinentalExchange, Inc. (ICE), its subsidiaries or affiliates, and does not constitute a contract or guarantee. Unencrypted electronic mail is not secure and the recipient of this message is expected to provide safeguards from viruses and pursue alternate means of communication where privacy or a binding message is desired.

This message may contain confidential information and is intended for specific recipients unless explicitly noted otherwise. If you have reason to believe you are not an intended recipient of this message, please delete it and notify the sender. This message may not represent the opinion of IntercontinentalExchange, Inc. (ICE), its subsidiaries or affiliates, and does not constitute a contract or guarantee. Unencrypted electronic mail is not secure and the recipient of this message is expected to provide safeguards from viruses and pursue alternate means of communication where privacy or a binding message is desired.

The code I showed would parse Smalltalk syntax comments from Smalltalk code. You are trying to parse class comments that were generated by a tool that I'm not familiar with. Those comments are not Smalltalk code; therefore, the code I sent is not useful to you except perhaps as something you can build on to recognize the comment syntax. You may find that the class comments were first generated by a tool using information that was discovered in the image; if so, you may be able to skip the class comment parsing and go directly to the source of those comments. If you can bypass the parsing of the comments then you'd eliminate the need to flog modelers for not keeping class comments "complete and correct"--which you say you have a way to check.

 

Sorry, I'm not familiar with XMI or what it is used for. To me it looks XML encoding with a specific set of tags.

 

Consider whether class comments are the best way to store class meta data. One way to store class meta data is with a VW syntax class definition that specifies attributes (look to #extraAttributesForDefinition for example). Another way, usually better, is to answer the meta-data by way of a class side method that can be queried and that is maintained like all other code. Here is an example of a method that returns class meta data used to describe observed attribute types:

 

icxMyAttributeTypes
 "Answer a collection of attribute names and bindings to observed types."
 ^#(
  (buttonHomes  ( #{Dictionary} ) )
  (currentView  ( #{HistoryView} ) )
  (viewDictionary  ( #{Dictionary} ) )
  (selectColor  ( #{ColorValue} ) )
  (unselectColor  ( #{ColorValue} ) )
  (selectedButton  ( #{ByteSymbol} ) )
 )

 

BTW, attribute type inference is not necessary when you have the ability to turn on a feature that "observes and revises" attribute types.

 

Paul Baumann 

 

---

From: Rob Vens [mailto:[hidden email]]
Sent: Monday, March 03, 2008 3:03 PM
To: Paul Baumann; VWNC
Subject: Re: [vwnc] parser for variable types in class comments
Importance: High

I am afraid I do not completely understand your point Paul. So let me explain a bit further to make sure you understood my problem. Please let me know if you did understood me correctly, and help me understand your point.
I am not trying to parse comments from XML, but from Smalltalk ... to use it in order to generate the correct XMI definitions. I am generating XMI (not XML) from Smalltalk, and the type information is not easily available in Smalltalk (except from parsing it from argument names like "aString" but that is far from robust - or inferring it from existing objects or messages sent to objects in code). So I am requiring the modellers to write the correct and complete class comments, which by the way I check for completeness and correctness before parsing it.
So for example this Smalltalk class:

Sogyo.Resilience defineClass: #Individual
    superclass: #{Core.Object}
    indexedType: #none
    private: false
    instanceVariableNames: 'name title gender addresses birth '
    classInstanceVariableNames: ''
    imports: ''
    category: ''

Which this class comment:
Natural persons.

Instance Variables:
    birth     <Sogyo.Resilience.Birth>
    addresses     <Set of: Sogyo.Resilience.PlatformAddress> the addresses of the person on the various defined platforms
    gender     <'male' | 'female'>
    title     <String> describing any titles
    name     <Sogyo.Resilience.PersonName>

This is exported to XMI as:

<UML:Class isRoot="false" name="Individual" visibility="public"
    xmi.id="Individual">
    <UML:ModelElement.taggedValue>
        <UML:TaggedValue tag="documentation"
            value="Natural persons.

Instance Variables:
    birth     &lt;Sogyo.Resilience.Birth>
    addresses     &lt;Set of: Sogyo.Resilience.PlatformAddress> the addresses of the person on the various defined platforms
    gender     &lt;'male' | 'female'>
    title     &lt;String> describing any titles
    name     &lt;Sogyo.Resilience.PersonName> " />
    </UML:ModelElement.taggedValue>
    <UML:Classifier.feature>
        <UML:Operation name="initialize" ownerScope="instance"
            visibility="public">
            <UML:ModelElement.taggedValue>
                <UML:TaggedValue tag="documentation"
                    value="initialize
   
    name := nil.
    addresses := Set new" />
            </UML:ModelElement.taggedValue>
        </UML:Operation>
        <UML:Operation name="name" ownerScope="instance"
            visibility="public">
            <UML:ModelElement.taggedValue>
                <UML:TaggedValue tag="documentation"
                    value="name
    ^name" />
            </UML:ModelElement.taggedValue>
        </UML:Operation>
        <UML:Operation name="name:" ownerScope="instance"
            visibility="public">
            <UML:ModelElement.taggedValue>
                <UML:TaggedValue tag="documentation"
                    value="name: anObject
    name := anObject" />
            </UML:ModelElement.taggedValue>
        </UML:Operation>
        <UML:Attribute name="title" ownerScope="instance"
            targetScope="instance" visibility="private">
            <UML:Attribute.initialValue>
                <UML:Expression />
            </UML:Attribute.initialValue>
            <UML:StructuralFeature.type>
                <UML:Classifier xmi.idref="SmalltalkDatatype00000002" />
            </UML:StructuralFeature.type>
            <UML:ModelElement.taggedValue>
                <UML:TaggedValue tag="description"
                    value="for description of title see class comment" />
            </UML:ModelElement.taggedValue>
        </UML:Attribute>
        <UML:Attribute name="gender" ownerScope="instance"
            targetScope="instance" visibility="private">
            <UML:Attribute.initialValue>
                <UML:Expression />
            </UML:Attribute.initialValue>
            <UML:StructuralFeature.type>
                <UML:Classifier xmi.idref="SmalltalkDatatype00000003" />
            </UML:StructuralFeature.type>
            <UML:ModelElement.taggedValue>
                <UML:TaggedValue tag="description"
                    value="for description of gender see class comment" />
            </UML:ModelElement.taggedValue>
        </UML:Attribute>
    </UML:Classifier.feature>
</UML:Class>

I am still working on the method argument XMI as you perhaps can see. Also I intend to implement a setting that will generate XMI using UML syntax for keyword messages, so that a method like name:address (with arguments aPersonName and anAddress) will be exported if desired as nameAddress(PersonName aPersonName, Address anAddress).

2008/3/3, Paul Baumann <[hidden email]>:

Rob,

 

You are trying to parse comments from XML, so this isn't directly relevant; however...

 

You could try a variation of the code in the attached file. The code is part of a scanner that efficiently compares code while ignoring formatting differences. I've only included the code relevant for extracting comments according to Smalltalk syntax. You might be able to customize for your parsing needs.

 

I'm surprised you need to parse the information you are looking for--instead of just generating it from code definitions.

 

Paul Baumann 

 

---

From: [hidden email] [mailto:[hidden email]] On Behalf Of Rob Vens
Sent: Sunday, March 02, 2008 5:00 AM
To: VWNC
Subject: [vwnc] parser for variable types in class comments

Has anybody done some work on parsing the class comments?
This uses a semi-formal notation, like:

Subclasses should implement:
accessing
   getPath

Instance Variables:
   name <String> the name of the thing
   employers <Set of: Person> a collection
   maritalStatus <'married' | 'single'>

etc.
For a tool I am writing to generate XMI exports from Smalltalk (currently for import into case tools, specifically Sparxsystems Enterprise Architect) I am using a brute force approach to parse this information, but possibly someone has done some work on this I am not aware of. I am not using the ParserCompiler toolkit or SmaCC for now.
The export to UML case tools is especially handy for eXploratory Modeling sessions, where Smalltalk is used to model the customers problem domain.

 

---

This message may contain confidential information and is intended for specific recipients unless explicitly noted otherwise. If you have reason to believe you are not an intended recipient of this message, please delete it and notify the sender. This message may not represent the opinion of IntercontinentalExchange, Inc. (ICE), its subsidiaries or affiliates, and does not constitute a contract or guarantee. Unencrypted electronic mail is not secure and the recipient of this message is expected to provide safeguards from viruses and pursue alternate means of communication where privacy or a binding message is desired.

 

---

This message may contain confidential information and is intended for specific recipients unless explicitly noted otherwise. If you have reason to believe you are not an intended recipient of this message, please delete it and notify the sender. This message may not represent the opinion of IntercontinentalExchange, Inc. (ICE), its subsidiaries or affiliates, and does not constitute a contract or guarantee. Unencrypted electronic mail is not secure and the recipient of this message is expected to provide safeguards from viruses and pursue alternate means of communication where privacy or a binding message is desired.