CPAN server for private cpan modules.
Well, it's very simple!
1. make repository mkdir -p repo-root/repo-one/ 2. copy distributives cp Foo-v0.0.1.tar.gz Bar-v0.0.2 repo-root/repo-one/ 3. create cpan index echo Foo-v0.0.1.tar.gz > repo-root/repo-one/packages.txt echo Bar-v0.0.2 repo-one >> repo-root/repo-one/packages.txt 3. start cpan repo server cpan_repo repo-root/ 4. now all the modules are available to install via cpan plus client! cpanp %cpanp /cs --add http://<cpan-repo-server>/repo-one/ %cpanp x %cpanp i Foo * <cpan-repo-server> - IP of CPAN::Repo server, if you run cpanp client on the same host where CPAN::Repo server run it could be 127.0.0.1
- simple maintenance - thanks to the author of CPANPLUS, custom sources idea is cool! for details see 'CUSTOM MODULE SOURCES' section in http://search.cpan.org/perldoc?CPANPLUS::Backend - multiply cpan repos accessible via one cpan server, see datails below - index merge, see datails below - no need to care about global cpan mirrors synchronization as with CPAN::Mini, because in cpanplus they exist separately (as custom sources)
As with example above only one repository with private cpan modules was addes. What if you want more? It's okay, just create as many as you wish, cpan repo server will care about further details:
1 making repositories and copy distribuitves mkdir -p repo-root/repo-two/ mkdir -p repo-root/repo-three/ cp Baz-v0.0.1 repo-root/repo-two/ cp Bazz-v0.0.1 repo-root/repo-three/ # generate index echo Baz-v0.0.1.tar.gz > repo-root/repo-two/packages.txt echo Bazz-v0.0.1 > repo-root/repo-three/packages.txt 2 re-setup cpanp client cpanp %cpanp /cs --remove http://<cpan-repo-server>/repo/ %cpanp /cs --add http://<cpan-repo-server>/repo-one/repo-two/repo-three/ %cpanp x 3 all three repos are available via cpan server! cpanp -i Foo cpanp -i Baz cpanp -i Bazz
The common approach with multiple repositories is to setup custom sources as:
where repo, repo2, repo-<i, ...> are names of repositories.
It's time to say about one great feature of CPAN::Repo called 'index merge'. Let's say we have module called 'Foo' of the version 'v0.0.2' in repository named 'repo-one' and module with the same name of the version 'v0.0.1' in another repository called 'repo-two', what happen if one add both repositories to cpan server as with following custom source url:
As it's seen from the url 'repo-one' is added first and 'repo-two' is added last. 'Index merge' means that repositorie's indexes added to cpan server are merged in order.
The modules from repository which added last will win and override versions of their "predecessors" . In the example above, one will get final version v0.0.1 of module Foo, all this logic is handled by cpan server and called 'index merge'.
'Index merge' is great possibility for:
* Delevloper's "sandbox" cpan repos which may be merged with "trunk" cpan repository during development cycle. * Repositories for patched modules coming from official cpan mirrors. * Repositoris with "frozen" versions of modules, allow to tag versions of your cpan modules and avoid installing last version of module when it's inadmissible. The main idea here is an isolation of your private cpan modules and capability to merge arbitrary cpan repositories during deploment process.
- repositories names should follow pattern /a-zA-Z_-/ - nested repositories are not supported, it means repo_root/foo/bar/ as repository place won't work, the same for distributives - they should be placed in the repository directory as plain list without subdirectroies - packages.txt - are custom source indexes should be kept actual and mantained by someone else, CPAN::Repo only provide read access for repository distributives and indexes and does proper index merge.
Alexey Melezhik / email@example.com