FAQ on CMT

Index

  1. Providing sourceless documents
  2. Accessing packages in a sub-directory

Providing sourceless documents

I'm always trying to use CMT together with java and I now have problems with the documentation.

What I'd like to do is just generating automatically my documentation using javadoc. Since it is not done in cmt, I tried to write a new fragment to define a new document type : javadoc.

The problem is that the process of generating the doc is not a file to file process : it does not take a .java file to write a .html. It takes a java package name (which has no associated file) and returns many files. This is not possible with the current implementation of document. It seems that adding this feature would lead to write some code into cmt_generator.

Does anyone see another method to generate my documentation ?

Sebastien

In fact what you ask for is already available in CMT, it is documented (although likely crypticly ;-) ) as follows :

Since your "document generator" does not make use of any source file per se, it's likely that you will provide as the fragments the following files (I use as an example the name "javadoc" for your document type):

...
make_fragment javadoc_header
make_fragment javadoc -header=javadoc_header
...
	

where javadoc is simply ... empty.

I remind you that here :

Then a document associated with the javadoc type could be specified as follows, using the unusual feature I mentioned above

document javadoc MyDoc JAVAPACKAGE=MyPackage
	

Here the syntax JAVAPACKAGE=MyPackage corresponds to providing user-defined templates in the document's fragments (see the general syntax) every variable name should then correspond within the fragment to constructs like (appearing anywhere in the fragment)

...
... ${JAVAPACKAGE} ...
...
	
(please note the ${..} syntax with required braces around the variable name)

You have probably already understood that CMT provides internal predefined similar variable names

the typical example being CONSTITUENT which is replaced by the constituent name when the fragment is instanciated.

What I'm talking about here consists in adding any number of user-defined variables which value will be specified on the document statement line using any number of <variable-name>=<value> constructs.

You will notice that the all-uppercase convention has been selected for CMT-internal variables (eg. CONSTITUENT). This is why I have used this convention for this example. This is only a convention, any form is possible here. The only constraint being that it should not overlap with a macro name or an env. var name.


Accessing packages in a sub-directory

I tried to use CMT with sub-packages. How does this work? If I have on cvs a structure like

 Velo/VeloEvent/..
     /VSicbCnv/..
	  

and on disk:

 Velo/VeloEvent/v3/..
     /VSicbCnv/v3/..
	  

it recognizes the stuff in the use statement, but when I say cmt show packages they are not there... Hence, the add-in for devstudio fails.

Well this is just a subtle effect :

Then in your precise case, you can do one of the following :

  1. If you strictly have (as you explained)
     xxx/Velo/VeloEvent/v3/..
             /VSicbCnv/v3/..
    	  

    then the CMTPATH should include xxx/Velo in order to make both VeloEvent and VSicbCnv visible.

  2. If Velo is itself a package :
     xxx/Velo/VeloEvent/v3/..
             /VSicbCnv/v3/..
             /v1/...
    	  

    then the CMTPATH may simply include xxx (although xxx/Velo is possible too but in this case, it is redundent). This makes all packages below Velo visible.

The difference between 1) and 2) is really the fact that in 2) Velo IS a true CMT package, whereas, in 1) Velo is just a prefix.

For simple efficiency reasons, CMT does not scan any arbitrary directory tree depth when it looks for all packages, but only the tree of true CMT packages + one level below each package.


Christian Arnault
Last modified: Thu Nov 9 15:16:35 MET 2000