global-localhost / package-name-map

A means of recording and referencing package names across ecosystems

package-name-map

A name-mapping database to help unify package managers. Every ecosystem makes different decisions for naming packages, and many also split up packages in different ways. To help package managers on any side of this wheel understand some other side of the wheel, we need to build a global map of equivalency between ecosystems.

Reads in a text file, currently in toml format, like so:

# packages from one ecosystem may map to many packages in another.  One entry should be present for each possible match
#  For example, here graphviz in conda is both a dev package (headers + link libraries) and a runtime, while Debian splits
#  out the dev package and the runtime.  Additionally, Debian is carrying an old name for the devel package, so we should
#  track that, too.
[[package]]
description="Runtime libraries and executables for graphviz"
conda="graphviz"
debian="graphviz"      # the runtime and executables

# Instead of duplicate entries for transitional or renamed packages, we collect lists
#   These could be extended into individual columns, or perhaps be one table per key that is then joined somehow
[[package]]
description="Development headers and libraries for graphviz"
conda="graphviz"
debian=["libgraphviz-dev", "graphviz-dev"]

CLI usage

This package installs an entry point, name-map which performs the desired functions.

To build the database from the toml input file,

name-map mkdb input_file.toml [database uri or filename]

Database URI is not required because a default sqlite database named package_name_map.db will be created in the current working directory if no other database URI is specified.

To use that database to output a JSON correspondence mapping between 2 package ecosystems, use the json subcommand:

name-map json pypi conda [-d database uri] [-o output json filename]

Here, database URI is optional again, because we look for .db files in the current working directory. If only one is found, we use it. If more than one is found, you'll get an error and a suggestion to specify the filename as an argument with the -d flag. For local sqlite databases, it is sufficient to provide a relative path to the file.