[dtp-dev] Query Builder initial code contribution

[Brian Payton <bpayton@xxxxxxxxxx>]

DTP dev community:

I've checked in an initial code contribution
for the Query Builder tool.  DTP CVS (/cvsroot/datatools) in the org.eclipse.datatools.sqltools/prototype/plugins
directory.  There are three plugins at this point:

  --  org.eclipse.datatools.sqltools.parsers.sql2003
  --  org.eclipse.datatools.sqltools.parsers.sql2003dml
  --  org.eclipse.datatools.sqltools.sqleditorx

The plugins are not currently part of
the DTP build process.  You can check them out from CVS into your
workspace and include them in your runtime environment if you want to play
with them.

This is *very* early code.  We're
preparing the foundation for the query builder, and our focus at this point
of development is on the interaction between the SQL parser and the SQL
sourece editor.  In fact these initial plugins don't include any "builder"(visual)
code at all.  

Plugin sqleditorx (a temporary name)
is based on the WTP SQL editor plugin (org.eclipse.wst.rdb.sqleditor).
 We started with the WTP SQL editor because it's simpler; we'll be
porting the code over to the DTP SQL editor base a little later.  Note:
I've temporarly disabled the connection-related code in the editor, so
you can write a query but not run it yet.

So what we current have is a new version
of the LPG-based parser and an editor that is tightly tied to it.  The
LPG-based parser uses a new technique for automatically generating the
parser AST (Abstract Syntax Tree) directly from the language grammar.  The
parser generator creates Java classes that correspond directly to the grammar
elements.  (This can significantly cut down the time needed to create
a new parser.)

In the editor we are trying to achieve
a user experience similar to the JDT Java editor.  That is, syntax
errors are diagnosed and marked as soon as you stop typing, and the error
messages are pretty good.  The LPG parser is actually two parsers
in one:  one parser that generates an AST, and another (the "diagnosing
parser")  that kicks in if the first parser finds an syntx error,
and tries to find and report as many errors as possible.   The outline
view is generated directly from the generated AST.  In the next drop
we should have full synchronization between the source view and outline
view (again, similar to the Java editor).  That is, if you click on
a location in the source view, the corresponding location in the outline
view is selected, and similarly when you click on a node in the outline
view, the corresponding element in the source view gets selected.

As I said, this is just a start.  There
are many more elements and features to add, including:
-- semantic validation (do tables and
columns exist?) using an associated SQL model
-- intelligent content assist based
on where you are in the statement
-- a parser manager for dynamically
selecting a parser based on the database connection type
-- code templates
and much more.  And that's just
for the source editor component.  We will be building the "visual"
builder around this foundation.  We have some cool ideas on how to
improve the interaction between the Assistant view (the "visual builder")
and the source view.  Stay tuned.

Brian Payton
SQL and XML Query Tools Development
IBM Silicon Valley Laboratory
(408) 463-2989, T/L 8-543-2989