aix27249 / agiliarepo Goto Github PK

There are some ideas to make abuilds-git prettier.

1. Each repo is single branch. Git remote & branch may be specified in repo object.
2. Abuilds with multiple packages: only main package name is real folder, all other package's folders is symlinks.
3. Some kind of subfolders, probably for each short tag. Current tree is ambiguous and hard to work with. On the other hand, real folders with abuilds may be somewhere else, keeps current structure, for example all tree may be moved to some subfolder and in the root directory additional shorttag-subfolders with symlinks added.

There are at least two possible models:

1. Each package record contains info about where it belongs (current implementation)
2. Each repository is a record itself, contains info about branches and subgroups, which stores a list of packages that belongs to it

Each approach has advantages and disadvantages.

Variant 1:

Advantages:

1. All data about package is stored as one object. It is easy to retrieve packages filtered by repo: result will contain data itself
2. If package is deleted, it automatically deletes from all repositories
3. No need for separate collection

Disadvantages:

1. Creating a repository clone requires update for all packages that belongs to it
2. If package belongs to different branches in different repositories simultaneously, it is difficult to store such data within that model

Variant 2:

Advantages:

1. Each repository is a separate object that contains package list only - easy to clone, easy to modify, easy to merge
2. Updates does not touch package data
3. Changes in different repositories are independent
4. Branch and subgroup data is related only to specific repository - still not very easy, but a little bit simpler

Disadvantages:

1. Retrieving package data itself requires two queries (or some kind of join)
2. If package is deleted, it is necessary to find them in every repository's branch and subgroup and delete it from there.

In some cases, it's necessary to display same block more than once - for example, in #content and in #sidebar. Since blocks can be configured to display different stuff depending on its position, this approach seems to be useful.

The problem is that each block is displayed as div id="block-$blockname", which causes duplicate IDs. Most browsers can workaround this problem, but it is not ok.

Need to find a proper way to distinguish these IDs. Keep in mind that CSS often appeals to block itself, and do that already in a #block-$blockname way.