On Tue, Oct 28, 2008 at 10:38 AM, Jean-Louis BOUDART <[EMAIL PROTECTED]> wrote: > 2008/10/27 Stefan Bodewig <[EMAIL PROTECTED]> >> I'm not good at names. I understand why this is a good name in the >> context of EasyAnt but struggle with the same name when applied to >> "pure" Ant. >> >> I recall being confused by the phase mapping when Xavier first >> suggested it.
> By typing ant checkstyle-report you only generate checkstyle report. > By typing ant javadoc you only generate javadoc report. > By typing ant report you generate ALL targets related to the report phase. > > This allow you more modularity than creating a target like : > <target name="report" depends="javadoc,checkstyle-report"/> > > Is it more clear? FWIW, I'm not fond of the "phase" name either. What you describe is better described as a target-group, with any target being able to join the group at will, and where there's no implicit intra-group dependencies, and no "body" to the target-group (i.e. the group is "abstract" in a way). Your example would read just as well, if not better, written as <target name="foo" group="bar" depends="...">...</target>. Just as you seem to propose, I'd be in favor of requiring your <phase> or my <target-group> to be explicitly declared before it can be referenced. This would avoid typos implicitly creating groups unbeknown to the build writer / fat fingered individual ;-) I think the notion of target-group is more self-explicit than the phase one. >> > - introduce two different kind of import (what we call use / >> > extends) >> >> I think you will quickly find support for this separation here, in >> particular if we throw in user definded prefixes where the writer of >> the build file that uses another build file gets to define the prefix >> of targets - hi, DD ;-) > > I really not understand what you mean here :'( Just Stefan pocking gentle fun at me, because I vehemently argued for what you propose when import was introduced. If you look back at the archives, I was strongly against using the import*ed* build file name in the import*er*, as it needlessly coupled the two, when good compartmentalization is essential to flexible and scalable designs (I even proposed the ability of the import*er* to give its own prefix as well). I also argued for separate import/include, just like in XSLT. In both cases my objections were ignored. So it's interesting that a separate group of Frenchmen now wants these same changes in Ant. When I argue what I believe is the right way to do something, and my point is being dismissed in favor of a lesser design (at least to me), this typically leaves me frustrated. Stefan "remue le couteau dans la plaie" comme on dit chez nous ;-) --DD --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
