It would be nice to have bio, matplotlib, and pandas narrowed down a bit. (numpy isn't needed, since it is a dependency of pandas)
Background info
At minimum, pinning to a major version (mydependency==2.*) is better than nothing (assuming the project follows semantic versioning). Pinning to an exact version is best for consistent behavior and reproducibility, but...
a problem with pinning versions (to a smaller range) is for people using the package, they might have a different version requirement for say pandas, which would cause install conflicts.
So the range of versions you want is a bit of a balance, and depends on the release strategy of the dependency. Dane, you know the bio package, so I leave that to you. Depending on how we are using pandas, it might be ok to pin it to a major version.
Can we get some version ranges for our dependencies? Right now we have
It would be nice to have
bio
,matplotlib
, andpandas
narrowed down a bit. (numpy isn't needed, since it is a dependency of pandas)Background info
At minimum, pinning to a major version (
mydependency==2.*
) is better than nothing (assuming the project follows semantic versioning). Pinning to an exact version is best for consistent behavior and reproducibility, but...a problem with pinning versions (to a smaller range) is for people using the package, they might have a different version requirement for say
pandas
, which would cause install conflicts.So the range of versions you want is a bit of a balance, and depends on the release strategy of the dependency. Dane, you know the
bio
package, so I leave that to you. Depending on how we are usingpandas
, it might be ok to pin it to a major version.