You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Organizations (JuliaML, JuliaStats,...) are the best candidates for producing best-of-breed bundles of packages related to their purpose. This need not be limited to the packages strictly owned by that organization. For example, JuliaML will want to include CatViews, Plots, etc even though those are not part of the org.
It should be dead-simple for an organization to maintain lists of bundles with various options. I propose allowing:
using MetaPkg
meta_add("JuliaML")
We will first check the MetaPkg/requires directory for a corresponding spec. If it's not found, we'll try to clone https://github.com/JuliaML/meta into MetaPkg/org/JuliaML.
The meta repo in an organization is a flat (non-julia) repo with only spec files (and maybe a license?). So the current MetaLearn file would be a file base.meta in that repo. Doing meta_add("JuliaML") would call meta_add on every spec file in that repo. There would be an option like meta_add("JuliaML","base") to only add that one spec file. The exception is that base would always be added.
Lets take the example of Plots. We could have a base bundle that includes just Plots itself, and a stats bundle that includes StatPlots, a glvisualize bundle that includes what is currently in MetaGL, etc. So one could choose to install specific bundles, or they could just do meta_add("JuliaPlots") and get everything in the ecosystem in one go.
The advantage of this design, aside from making it super easy to get set up, is that we would have a well-maintained collection of curated package lists with categorical groupings. We could presumably make it easy to download or browse the spec trees, and not too hard to automate building a site like https://github.com/svaksha/Julia.jl.
Additionally, one could easily build their own meta-bundles (custom standard libraries) just by including other meta repos in their spec. JuliaLang could (should) have bundles to include JuliaMath, etc, so that one could easily mix and match their preferred standard library, or just get everything all at once.
The text was updated successfully, but these errors were encountered:
Cool! I wonder what the pros and cons for a straight forward package would be. I guess you started off like this, because it was mostly about checking out different branches of a bunch of packages? @timholy this and #6 should be interesting to you, too ;)
Organizations (JuliaML, JuliaStats,...) are the best candidates for producing best-of-breed bundles of packages related to their purpose. This need not be limited to the packages strictly owned by that organization. For example, JuliaML will want to include CatViews, Plots, etc even though those are not part of the org.
It should be dead-simple for an organization to maintain lists of bundles with various options. I propose allowing:
We will first check the
MetaPkg/requires
directory for a corresponding spec. If it's not found, we'll try to clonehttps://github.com/JuliaML/meta
intoMetaPkg/org/JuliaML
.The
meta
repo in an organization is a flat (non-julia) repo with only spec files (and maybe a license?). So the currentMetaLearn
file would be a filebase.meta
in that repo. Doingmeta_add("JuliaML")
would callmeta_add
on every spec file in that repo. There would be an option likemeta_add("JuliaML","base")
to only add that one spec file. The exception is thatbase
would always be added.Lets take the example of Plots. We could have a
base
bundle that includes just Plots itself, and astats
bundle that includes StatPlots, aglvisualize
bundle that includes what is currently inMetaGL
, etc. So one could choose to install specific bundles, or they could just dometa_add("JuliaPlots")
and get everything in the ecosystem in one go.The advantage of this design, aside from making it super easy to get set up, is that we would have a well-maintained collection of curated package lists with categorical groupings. We could presumably make it easy to download or browse the spec trees, and not too hard to automate building a site like https://github.com/svaksha/Julia.jl.
Additionally, one could easily build their own meta-bundles (custom standard libraries) just by including other meta repos in their spec. JuliaLang could (should) have bundles to include JuliaMath, etc, so that one could easily mix and match their preferred standard library, or just get everything all at once.
The text was updated successfully, but these errors were encountered: