Add support for BYTEA/BLOB #511
Conversation
|
@JelteF currently the output from |
| \x | ||
| \x5c7831315c7830315c7830325c7830335c7830345c7830355c7830365c7830375c7830385c7830395c7830415c7830425c7830435c7830445c7830455c783046 | ||
| \x | ||
| \x5c783030 |
There was a problem hiding this comment.
currently the output from select * from blob_tbl is an escaped version of inserted rows, is this fine or do we need to change that?
Okay, this needs to change. All the non empty bytea results are wrong. When using postgres execution the it instead gives the following rows:
\x
\x110102030405060708090a0b0c0d0e0f
\x
\x00
\x07
It seems like you're encoding the string as hex somewhere as an additional time, because 5c783030 is \x00 in ASCII, i.e. including the \ and x characters. So I think when converting between from DuckDB to PG type you're encoding the string representation of the type instead of the raw bytes.
There was a problem hiding this comment.
Found it, converting it to string was trying to convert those bytes to string
There was a problem hiding this comment.
Changed GetValue to GetValueUnsafe for getting raw bytes in string_t
| \x07 | ||
|
|
||
| (6 rows) | ||
|
|
There was a problem hiding this comment.
Can you add a comparison test too? Something like:
SELECT * FROM blob_tbl WHERE a = '\x00';
| auto str = value.GetValueUnsafe<duckdb::string_t>(); | ||
| auto blob_len = str.GetSize(); | ||
| auto blob = str.GetDataUnsafe(); | ||
| bytea* result = (bytea *)palloc0(blob_len + VARHDRSZ); |
There was a problem hiding this comment.
nit: make format will re-format this line as
bytea *result = (bytea *)palloc0(blob_len + VARHDRSZ);
There was a problem hiding this comment.
Thanks for the report. Fixed by: #518
In #496 we started using the newest clang-format to format our files, but CI was still installing the old version. This meant that we were not catching some unformatted files correctly. An example of this being: #511 (comment) This starts using the correct clang-format version in CI too and formats any incorrectly formatted files.
In #496 we started using the newest clang-format to format our files, but CI was still installing the old version. This meant that we were not catching some unformatted files correctly. An example of this being: #511 (comment) This starts using the correct clang-format version in CI too and formats any incorrectly formatted files.
In duckdb#496 we started using the newest clang-format to format our files, but CI was still installing the old version. This meant that we were not catching some unformatted files correctly. An example of this being: duckdb#511 (comment) This starts using the correct clang-format version in CI too and formats any incorrectly formatted files.
Fixes #464