----- Mail original ----- > De: "Peter Levart" <peter.lev...@gmail.com> > À: "Jochen Theodorou" <blackd...@gmx.org>, jigsaw-dev@openjdk.java.net > Envoyé: Mercredi 2 Décembre 2015 17:45:59 > Objet: Re: using modules for caching > > > > On 12/02/2015 01:40 PM, Jochen Theodorou wrote: > > Hi, > > > > I was wondering if it will be possible to reuse module information between > > runs.. Currently I am collecting a lot of reflective information and use > > it to build a meta class system, which I then need to do method > > invocations, what fields are available and other things like that. Since > > that can take a while I was wondering if it is possible to cache this > > information which made me wonder if there is something that allows me to > > interact with modules to know if their composition has changed compared to > > the last run. For example that all classes in module X are still the same. > > I was thinking that the JVM itself would use this kind of information for > > runtime images, thus maybe there are ways to get this kind of information? > > > > bye Jochen > > An idea: > > You collect this information and generate a special module from it. Then > you build a runtime image composed of modules of original image and that > additional module. This could be performed by a special variant of jlink > - a glink!
You don't need to collect something before calling jlink, jlink has a system of backends that you can plug to analyze the runtime image and produce what you want (as far as i know the interface of a backend is not settle). Trying to do analysis like this at compile time is maybe not a good idea because modules can change between compile time and install time. > > Peter > cheers, Rémi
