Hi John,
The document doesn't exactly explain what you were doing other than
swapping in plexus-interpolation. Is that effectively the gist of it?
Better isolating the code without making any functional changes so it
is more stable, while the other proposal was for fixing something?
The code looks good to me. The devil is always in the details and I'm
happy with it as long as we've battle tested it enough. It won't hit
all the cases, but I was thinking I could extend my tool that resolves
all the POMs in a repository (tried against central) and see if this
yields different results with the new interp code. WDYT?
Small comment on the RegexBasedModelInterpolator - the static block
could be replaced with an array initializer like you have above it for
the other var :)
With the build timestamp addition, I'm not sure if I like that the
format is specified by an arbitrary property in the POM? Just a
thought really - I would hope plugins can access the variable for the
actual date and use it more than being able to substitute it into the
POM (bear in mind two different uses might want different formats in
the same POM). Possibly getting a bit "out there", but what if we
could paramaterise interpolation for certain formats instead? eg. $
{maven.build.timestamp|format='yyyyMMdd'}
Cheers,
Brett
On 17/06/2008, at 3:35 AM, John Casey wrote:
Hi everyone,
I wanted to start another discussion having to do with POM
interpolation in the upcoming Maven 2.0.10 release. If you recall, I
previously sent out a proposal for retaining dynamic expressions in
the build section of the POM, to allow updates to that section that
take place via plugin execution to be reflected in subsequent
executions.
This proposal is somewhat different, and is mostly orthogonal to
that previous discussion. This proposal is less about _when_ we
interpolate different parts of the POM, and more about _how_ we do
it. Since before its original release, Maven has used a single-class
implementation of the ModelInterpolator used to resolve expressions
in the POM. As we learned more about what we needed to support in
POM expressions - and, what the order of operations was for
interpolating - this single class quickly grew into quite a beast of
tangled conditional logic and loops. Even the interface and the code
surrounding the call to that interface were abused more than a
little, as we tried to trick the interpolator into making exceptions
of certain parts of the POM, and prioritizing certain sources of
expression values over others.
The full proposal is here:
http://docs.codehaus.org/display/MAVEN/POM+Interpolation+Refactor
As a brief synopsis:
For anyone who has tried to improve the interpolator ahead of a
2.0.x release, you already know what I'm talking about; the
interpolator has been limping along for some time now. For anyone
who hasn't spent much time on this functionality, a quick scan of
JIRA for any given release will show you how persistent and varied
the interpolation problems have been. It has reached the point where
solving these problems within the current implementation often
requires threading your way through rules that seem diametrically
opposed to one another.
At the same time, I've been working on the interpolation code in
plexus-utils for use in the Assembly plugin (and though I'm not sure
who you are, I know others are using plexus-utils for interpolation
as well). I believe this code has reached a level of maturity and
sophistication that will allow us to replace the original
interpolator in maven-project with a less-complex solution
implemented around this plexus-utils code. However, because of the
historic problems with plexus-utils in Maven, and because plexus-
utils contains several other hotspots of variability like the CLI
code used to fork child processes, I split this interpolation code
out into a new library called plexus-interpolation. The old plexus-
utils classes now depend on plexus-interpolation, which is shaded
into plexus-utils until those deprecated classes can be removed.
Additionally, I've implemented this refactor on a feature branch, so
you can take a look at:
https://svn.apache.org/repos/asf/maven/components/branches/john-2.0.x-plxInterpolation
to see the actual implementation.
Please take a look and tell me what you think.
Thanks,
-john
--
John Casey
Developer, PMC Member - Apache Maven (http://maven.apache.org)
Blog: http://www.ejlife.net/blogs/buildchimp/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Brett Porter
[EMAIL PROTECTED]
http://blogs.exist.com/bporter/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
