v0.19.0

Removing variable attributes

filter_cdf(variable_attributes=...) keeps the variable attributes that match, like attributes= does for global ones.

• Names or a pattern apply to every variable.
• A function gets the variable and the attribute, to target one variable.
• With add_attribute, it renames an attribute:

frame = cdf["B"].attributes["Frame"]
new = cdf.filter(variable_attributes=lambda v, a: not (v.name == "B" and a.name == "Frame"))
new["B"].add_attribute("FRAME", frame.value, frame.type())

Held variables and attributes stay valid

A variable or attribute held from Python now stays valid when other entries are added or removed. Before, it could read freed memory, or another entry, and crash.

• Removing a variable or an attribute keeps the order of the others. It used to move the last one into its place, which also changed the order in saved files.
• Only a wrapper of the removed entry itself dangles.

CI

A stalled download of the test corpus no longer hangs a job: downloads and wheel tests time out.