gobo-eiffel-develop

[gobo-eiffel-develop] ET_AST_ITERATOR.process_features

[Howard T. <how...@di...>]

Hi Eric,

In working on my Eiffel comparison code, I found that the process_features feature of ET_AST_ITERATOR
fails to process features in some classes.

According to feedback from my EDP project, this feature is not part of the reachable feature set of the Eiffel
compilation process in 'gec' and is therfeore untested in that environment.

I append a reimplementation that seems to work somewhat better, although not yet proven correct:

	process_features (a_class: ET_CLASS) is
			-- Process feature clauses of `a_class'.
		require
			a_class_not_void: a_class /= Void
		local
			a_feature_clauses: ET_FEATURE_CLAUSE_LIST
			a_feature_clause: ET_FEATURE_CLAUSE
			l_queries: ET_QUERY_LIST
			l_query: ET_QUERY
			l_procedures: ET_PROCEDURE_LIST
			l_procedure: ET_PROCEDURE
			i, nb: INTEGER
			j, nb_queries: INTEGER
			k, nb_procedures: INTEGER
		do
			a_feature_clauses := a_class.feature_clauses
			if a_feature_clauses /= Void then
				from
						-- Setup for 'feature' sequence
					i := 1
					nb := a_feature_clauses.count
						-- Setup for 'queries' sequence
					j := 1
					l_queries := a_class.queries
					nb_queries := l_queries.count
					if j <= nb_queries then
						l_query := l_queries.item (j)
					else
						l_query := Void
					end
						-- Setup for 'procedures' sequence
					k := 1
					l_procedures := a_class.procedures
					nb_procedures := l_procedures.count
					if k <= nb_procedures then
						l_procedure := l_procedures.item (k)
					else
						l_procedure := Void
					end
				until
					i > nb
				loop
						-- Set feature clause for this iteration
					a_feature_clause := a_feature_clauses.item (i)

						-- Process either the next query, or the next procedure
						-- and select the next query or procedure ...
					if l_query /= Void and then l_query.feature_clause = a_feature_clause then
						if l_procedure /= Void and then l_procedure.feature_clause = a_feature_clause then
							if l_query.name.position < l_procedure.name.position then
									-- Here, both the current query and the current procedure lie within
									-- the current feature, and the query comes first
								l_query.process (Current)
								j := j + 1
								if j <= nb_queries then
									l_query := l_queries.item (j)
								else
									l_query := Void
								end
							else
									-- Here, both the current query and the current procedure lie within
									-- the current feature, and the procedure comes first
								l_procedure.process (Current)
								k := k + 1
								if k <= nb_procedures then
									l_procedure := l_procedures.item (k)
								else
									l_procedure := Void
								end
							end -- if
						else
								-- Here, only the query lies in the current feature clause
							l_query.process (Current)
							j := j + 1
							if j <= nb_queries then
								l_query := l_queries.item (j)
							else
								l_query := Void
							end
						end -- if
					elseif l_procedure /= Void and then l_procedure.feature_clause = a_feature_clause then
							-- Here, only the current procedure lies within the current feature clause
						l_procedure.process (Current)
						k := k + 1
						if k <= nb_procedures then
							l_procedure := l_procedures.item (k)
						else
							l_procedure := Void
						end
					end -- if
					
						-- If both the current query and procedure are not in
						-- the current feature clause, then select the next feature clause
					if (l_query = Void or else l_query.feature_clause /= a_feature_clause)
					and then (l_procedure = Void or else l_procedure.feature_clause /= a_feature_clause) then
						i := i + 1
					end -- if
				end -- loop
			end -- if
		end

Cheers,

Howard Thomson

By the way, did you get my previous e-mail about waiting for me to update SF svn ?

--
"Only two things are infinite, the universe and human stupidity,
and I'm not sure about the former." -- Albert Einstein 

Re: [gobo-eiffel-develop] ET_AST_ITERATOR.process_features

[Eric B. <er...@go...>]

Howard Thomson wrote:
> In working on my Eiffel comparison code, I found that the process_features feature of ET_AST_ITERATOR
> fails to process features in some classes.
> 
> According to feedback from my EDP project, this feature is not part of the reachable feature set of the Eiffel
> compilation process in 'gec' and is therfeore untested in that environment.

It is in fact tested by $GOBO/test/tools.

> I append a reimplementation that seems to work somewhat better, although not yet proven correct:

It is in fact proven not correct when running the test $GOBO/test/tools.

Can you be more explicit about what ET_AST_ITERATOR.process_features
fails to process? Is your AST well formed with synonyms correctly
chained?

> By the way, did you get my previous e-mail about waiting for me to update SF svn ?

Yes.

-- 
Eric Bezault
mailto:er...@go...
http://www.gobosoft.com

Re: [gobo-eiffel-develop] ET_AST_ITERATOR.process_features

[Howard T. <how...@di...>]

Hi Eric,

It doesn't work, as you say !

My gecmp program uses Gobo tools to parse pairs of class files, uses a 
derivation of ET_AST_PRINTER to accumulate the sequence of ET_AST_LEAF 
elements that were parsed, does a 'diff' pass over the pair of sequences 
(arrays) of leaf elements, marks the elements according to whether they are 
equal, not-equal, or excluded, and uses a derivation of ET_AST_PRINTER to 
print the sequence of elements tagged with the "== ","/= ","   " marking.

The process worked fine on my initial test pair of classes with just a single 
feature clause and one feature.

When I started processing all classes in the Universe for the EDP project, 
modified to add the current $GOBO/library/tools/eiffel class set to provide 
pairs of class files for comparison, I ended up with (some) classes with 
feature clauses with only the word 'feature' and no features at all.

I don't yet fully understand the implications of the presence of synonyms in a 
class, and I am unsure as to what assumptions one can make about the ordering 
of features in ET_CLASS.queries and ET_CLASS.procedures.

I need it to work for all valid class texts, so I will find the problem, 
eventually !

Howard

On Wednesday 22 August 2007 08:29, you wrote:
> Howard Thomson wrote:
> > In working on my Eiffel comparison code, I found that the
> > process_features feature of ET_AST_ITERATOR fails to process features in
> > some classes.
> >
> > According to feedback from my EDP project, this feature is not part of
> > the reachable feature set of the Eiffel compilation process in 'gec' and
> > is therfeore untested in that environment.
>
> It is in fact tested by $GOBO/test/tools.
>
> > I append a reimplementation that seems to work somewhat better, although
> > not yet proven correct:
>
> It is in fact proven not correct when running the test $GOBO/test/tools.
>
> Can you be more explicit about what ET_AST_ITERATOR.process_features
> fails to process? Is your AST well formed with synonyms correctly
> chained?
>
> > By the way, did you get my previous e-mail about waiting for me to update
> > SF svn ?
>
> Yes.

-- 
Howard Thomson

--
"Only two things are infinite, the universe and human stupidity,
and I'm not sure about the former." -- Albert Einstein 

Re: [gobo-eiffel-develop] ET_AST_ITERATOR.process_features

[Eric B. <er...@go...>]

Howard Thomson wrote:
> I don't yet fully understand the implications of the presence of synonyms in a 
> class,

The trick is that the ET_AST_PRINTER prints:

    foo, bar (...) is
        do
           ...
        end

and not:

    foo (...) is
        do
           ...
        end

    bar (...) is
        do
           ...
        end

when it finds synonyms. This is needed when what we want it just
change the formating of a class text without decoupling synonym
features. This trick about synonyms is done if class ET_AST_ITERATOR
in feature `process_features' and for example in feature
`process_do_procedure'. If you don't want this behavior and want
the synonym features to stand by themsleves, then you would need
to redefine these features in a descendant of ET_AST_PRINTER.

> and I am unsure as to what assumptions one can make about the ordering 
> of features in ET_CLASS.queries and ET_CLASS.procedures.

Queries in ET_CLASS.queries between 1 and `queries.declared_count' are
guaranteed to be sorted in the order they appear in the class text.
Likewise for procedures. However we know nothing about the ordering
between a query and a procedure. We have to compare their position
in the class text in that case.

-- 
Eric Bezault
mailto:er...@go...
http://www.gobosoft.com