Add codec for "aclitem[]" type #58
Conversation
Previously the "aclitem" type specified to be transferred in the text mode because Postgres does not support sending it in the binary mode, but the "aclitem[]" did not have a support at all.
|
Tests? Why do we need 90806d3 as part of this PR? |
I've just added them, no failures (but I fixed a bug, thank you).
Because ac0dfe1 is based on it. If you want to use it separately from this PR, you can move master to this commit (or 3e4e3f1) or even cherry-pick it (them), I'll rebase the other ones. |
vitaly-burovoy
force-pushed
the
aclitem
branch
2 times, most recently
from
d969f90 to
ac0dfe1
Compare
|
IMO we should generalize this into a common array literal parsing function and just add text codecs for types that do not support binary I/O. Ditto for composites and ranges. |
|
+1 to what @elprans suggests. |
|
I merged the first half of this PR. Generic support for text arrays is tracked separately in #60. Thanks! |
|
Thank you very much! |
The "aclitem[]" type can't be fetched without a cast. Moreover codec for it can't be (re-)defined from a Python code.
The last commit fixes that by adding a codec.
The second one makes error message (for similar cases) be the same (helped during debugging).
The first one stabilizes generating of "pgtypes.pxi" (to make the last commit be less complex).