I've spent a little time over the last week trying again to tease apart GWT
into smaller modules and working out dependencies between them. This is in
part to see about a not-ant build system, and in part to see why a minimal
GWT app is surprisingly big. The target here isn't the user lib, but the
compiler and associated other pieces that are all compiled together:
emulation, internal annotations, the generator/linker api, commandline
tools, testing infrastructure, etc.
Along the way, the biggest snarl I've run into so far is Util, and its
siblings, Utility and StringUtils. "A smattering of useful methods" indeed.
Utility has the benefit of depending on very little outside the JRE, but
gets used all over the place (in Util, the compiler, dev mode, linkers,
generators, command-line tools...) - not such a big deal by itself, just
need to make sure it is in one of the core modules so that everything else
can lean on it. StringUtils likewise depends on nothing and is used all
over with one additional usecase: servlets.
Util on the other hand depends on the generator types
(treelogger, UnableToCompleteException), on SpeedTracer (what is the status
of SpeedTracer these days?), and on the org.w3c.dom library. It also gets
used pretty much everywhere, which adds transitive dependencies to the
above: the compiler, dev mode, generators, linkers, command line tools. It
also has this confusing Javadoc:
> Methods in this class are candidates for being moved to {@link
com.google.gwt.util.tools.Utility} if they would be generally useful to
tool writers, and don't involve TreeLogger.
Only six methods (out of dozens) take either the logger or throw that
UnableToCompleteException
I'd tend to say that all generators shouldn't depend on org.w3c.dom or
SpeedTracer to be on their classpath just to compile including any
ResourceGenerator for ClientBundle (and yes, admittedly, Generators and
Linkers are going away when GWT3 starts showing up, but until then...), the
Link.main() likely doesn't need to depend on generators, etc.
To that end, I'm looking at teasing apart some of these into distinct
classes, purpose-built for where they are used, and only with the
dependencies they need. Any objections to this that would prevent it
getting merged into master?
It is just over 1000 lines long (plus another 350 for Utility), and the
first cut appears to be taking out org.w3c.dom. This slices out just over
10%, and at no real cost that I've noticed yet - all the methods were
private, and at least ant dist-dev passes without them, I'll be running a
longer build soon.
Thoughts?
--
You received this message because you are subscribed to the Google Groups "GWT
Contributors" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/google-web-toolkit-contributors/617fd0e4-1604-458e-88fe-da686ab6d31d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
