Documentation
The documentation for the dbm module, eg at https://docs.python.org/3.13/library/dbm.html include an example storing keys and values which are of type str. The documentation says about strings:
Key and values are always stored as [bytes](https://docs.python.org/3.13/library/stdtypes.html#bytes).
This means that when strings are used they are implicitly converted
to the default encoding before being stored.
It is not at all clear to me what "the default encoding" means. For example, one might assume it was the encoding from locale.getenconding() but I think not. Looking at the dbm.sqlite code one sees CAST(? AS BLOB) as the insert parameter placeholder. That says to me that the encoding is whatever the database is using, and that is not apparent to me from looking at the code. And I imagine that the other dbm backends may use other, different, choices for the default encoding.
Ideally I'd like:
- an explicit statement about how the encoding is chosen if that is possible, or a statement that this is backend and possibly current-locale dependent otherwise
- a statement that users should probably always do their own conversion to bytes before storing values if they want control
- possibly an optional
encoding parameter for the dbm.open calls to provide an encoding, with the current (vague but historicly compatible) behaviour if unset, but if provided then an explicit catch of str values in __setitem__ if set, and conversion according to what was provided
I can probably make a PR for the second and third items, and the "backend dependent" flavour of the first one.
Linked PRs
Documentation
The documentation for the dbm module, eg at https://docs.python.org/3.13/library/dbm.html include an example storing keys and values which are of type
str. The documentation says about strings:It is not at all clear to me what "the default encoding" means. For example, one might assume it was the encoding from
locale.getenconding()but I think not. Looking at thedbm.sqlitecode one seesCAST(? AS BLOB)as the insert parameter placeholder. That says to me that the encoding is whatever the database is using, and that is not apparent to me from looking at the code. And I imagine that the other dbm backends may use other, different, choices for the default encoding.Ideally I'd like:
encodingparameter for the dbm.open calls to provide an encoding, with the current (vague but historicly compatible) behaviour if unset, but if provided then an explicit catch ofstrvalues in__setitem__if set, and conversion according to what was providedI can probably make a PR for the second and third items, and the "backend dependent" flavour of the first one.
Linked PRs