Re: [relax-ng] trang instance input module

[James Clark <jjc@jclark.com>]

> An instance input module for Trang would be a nice feature. That is,
> allowing Trang to take an XML document as input and then best-guessing a
> schema from it. From a schema developer's point of view, this feature
> would be very good to have.

I think we can distinguish a couple of scenarios:

a) A user already has a collection of real example documents available. 
Perhaps they have been getting by without a schema, but have now decided it 
would be nice to have a schema for their document format.

b) A user is creating the example instance as part of the process of 
designing the document format.  They are using an example rather than 
writing a schema directly because they find it easier.

In case (b), I think Examplotron makes good sense.

In case (a), something else is needed, more like Microsoft's XSD inference 
tool.  The big question in my mind is can this be done well enough to be 
useful without some manual guidance from the user.  Is it better to 
generate something loose (and consequently smaller) and let the user 
tighten it up, or something very constraining (and perhaps rather large) 
and let the user generalize it?

Maybe an interactive tool would be more useful in this case. The tool could 
present intelligent choices for the user to make with sensible defaults 
based on the instance documents.  For example,

- the user could specify that a particular set of elements share a common 
content model

- for a particular attribute, there may be multiple datatypes that match 
all the values found; the tool would allow the user to choose which of 
these would be used

- an attribute of a particular name might be found on elements with 
different names; the tool would allow the user to say whether the value of 
the attribute depended on the parent element or not

Do any of the exisiting commercial tools do this sort of thing?

James