[pdal] Conda for official Mac/Windows PDAL packages?

Kristian Evers kreve at sdfe.dk
Fri Jun 1 06:18:01 PDT 2018


Howard,

As a user of both OS X, Windows and occasionally PDAL, I would welcome such a change. As a user I find both Homebrew and OSGeo4W less than optimal. Homebrew is generally a nice environment but it has proven difficult to maintain the various OSGeo packages in a way that is easily compatible with the core homebrew set of packages.
As a developer/maintainer I unfortunately find OSGeo4W very fragile at the moment. Primarily because of the few people who are actively maintaining it (it is more or less a one person operation it seems). On top of that it is very difficult to be a "drive-by" contributor because ultimately SSH access is required. Homebrew works fairly well from a maintainer perspective, but when it comes to the OSGeo packages it is still problematic.

The conda forge ecosystem removes all of the above problems and in my experience everything just works. Both from a user and maintainer perspective. So for what it's worth, a +1 from me.

/Kristian

Fra: pdal [mailto:pdal-bounces at lists.osgeo.org] På vegne af Howard Butler
Sendt: 1. juni 2018 14:53
Til: pdal <pdal at lists.osgeo.org>
Emne: [pdal] Conda for official Mac/Windows PDAL packages?

All,

Brad's recently announced [1] Conda Forge packages for PDAL proved wildly successful for our recent FOSS4GNA workshop [2]. The current PDAL packaging story is rather messy, with OSGeo4W (64-bit only) available for Windows, Homebrew for OSX, and Debian/RedHat packages being provided by the distributions. Docker is also in the mix, although that's a separate thread

The OSGeo4W package creation process is now fully automated through PDAL's AppVeyor continuous integration, but the upload, tagging, and release of the package is still a manual activity. OSGeo4W can also be very brittle for people, especially as an installation, and it often lacks the refresh of underlying system support packages on which everything is based.

Homebrew PDAL packages for OSX are also available, but they are only available to 1.6 and installations can be fickle, especially when it comes to updating.

The appeal of Conda Forge as a packager is a single upstream for packaging efforts, a build system that doesn't require our project to maintain it (currently done with OSGeo4W), and convenient usability of the packages for the users.  Another strong appeal of Conda Forge is the workflow of package creation, changes, patches, and refresh can be done through typical pull-request workflow. OSGeo4W updates require SSH access to a machine (that is sometimes down), and manual button mashing with careful manual post validation.

I'm proposing that we drop the OSGeo4W and Homebrew package support in exchange for the Conda Forge ecosystem. I'm interested in hearing from users of the Homebrew and OSGeo4W packages on that topic, however. Would this drastically upend your workflow? Make things simpler? If these packaging approaches are critical for you, can you step forward to take over their care with some hand-off?

If there's no significant objection, PDAL 1.8 will target Conda as the primary packaging environment for Mac/Windows.

Howard



[1] https://lists.osgeo.org/pipermail/pdal/2018-April/001559.html
[2] https://2018.foss4g-na.org/session/point-cloud-processing-and-analysis-pdal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/pdal/attachments/20180601/b173c2f6/attachment.html>


More information about the pdal mailing list