- rust: cargo
- NodeJS: npm
- D: dub
- fortran: fpm
- etc…
Would that be something which would be of interest for a PB v6?
Yes, I am aware of that. From the beginning, the CodeArchiv has the role of archiving the PB source codes from the forums in a categorized form where all the source codes work out-of-the-box. No cumbersome copy and paste of the source codes from the forum posts.Yann64 wrote:I am not too sure the package sources should all be hosted in the same git repository as I fear managing package versions, package specific tickets, will quickly become too difficult as you will end-up with a huge repository.
Yes, this is how I have planned it. There will be a new repository "PB-Packages", which will then only provide a package list and the source codes in the CodeArchiv will each be moved to separate repositories as packages. The authors of the packages should create their own repositories and manage them themselves.Yann64 wrote:I would rather have a main repository somehow managing the package list. Packages themselves would be hosted in their own repository (structure to be standardized), with the added benefits that each of them can be imported independently (no git clone of the entire repository)
Are you talking about PB PureLibraries? As I can read from this README (purebasic/sdk/libraries-maker/README), the creation of these PB PureLibraries requires that the source code is written in ASM or C. However, most write PB source codes here, so this is not an option. But recently it was announced by the PB developers that they are developing a C compiler backend for the PB compiler. Maybe then PB PureLibraries can also be created with PB source codes. I am curious.Yann64 wrote:compiled as a library for direct linking at compile time
Not required for PB PureLibraries as far as I know. And in case the library is PB source code, the package manager can create such a catalog automatically during package installation.Yann64 wrote:and even host their own catalog to enable auto-completion when the library is imported
Maybe we could extract the PB source code of this function from the PureBasic IDE source code and adapt it for this purpose.Scan file for AutoComplete
Files with this option will be scanned for AutoComplete data even when they are not currently loaded in the IDE. This option is on by default for all non-binary files. It should be turned off for all files that do not contain source code as well as for any files where you do not want the items to turn up in the AutoComplete list.
Yes, definitely. There are already discussions about this and some other topics: Improvement of code metadataYann64 wrote:There should probably be a definition file at the root of the repository to define dependencies to other packages.
The solution should not be PB IDE dependent, because PB source codes can also be compiled without the PB IDE. As written above, unfortunately I haven't found a good solution yet to include PB source codes automatically.Yann64 wrote:As the IDE was open sourced, it would maybe be possible to pass flags as compile time pointing to the downloaded libs?
Yes, should work as a PB IDE tool, but also as a standalone tool.Yann64 wrote:The package manager IDE can probably simply be added in the Tools menu (for a start).
Sure. I will do that. But currently I think I'll wait until there are clean ways to be able to include packages automatically so that auto-complete is also available. Developing the package manager and everything that goes with it will be a lot of work again. Because of the mentioned reasons, I see therefore currently no sense to start the work. As written above, it may be possible soon when the C compiler backend is developed and then maybe PB PureLibraries can also be built from PB source code.Yann64 wrote:For source code management solutions (git, mercurial, ...) to be used by the community, I guess someone will have to come-up with a clear benefit for the community to start adopting it. A package would bring that benefit, so I guess someone has to deliver to initiate the whole thing