Thread: Using composite types in psycopg3
Hello,
I have tried psycopg3 very briefly and I have a question.
I have a couple of use cases in systems I currently work with
that prevented (or seriously limited) usage of psycopg2, so
I had to use other drivers. This generally relates to accessing
composite types that you can construct on-the-fly in SQL queries.
Consider the following example:
'''
import psycopg2
def get_query(fpath):
with open(fpath, 'rt') as f:
return f.read()
def main():
conn = psycopg2.connect('postgres://user:password@host/db')
cur = conn.cursor()
cur.execute(get_query('query.sql'))
result = cur.fetchone()[0]
print(type(result), result)
cur.close()
conn.close()
if __name__ == '__main__':
main()
def get_query(fpath):
with open(fpath, 'rt') as f:
return f.read()
def main():
conn = psycopg2.connect('postgres://user:password@host/db')
cur = conn.cursor()
cur.execute(get_query('query.sql'))
result = cur.fetchone()[0]
print(type(result), result)
cur.close()
conn.close()
if __name__ == '__main__':
main()
'''
Where query.sql is:
'''
with test as (
select 1 as id, 'one' val
union all
select 1, 'one more'
union all
select 2, 'two'
)
select array(
select (id, array_agg(val))
from test
group by id
)
select 1 as id, 'one' val
union all
select 1, 'one more'
union all
select 2, 'two'
)
select array(
select (id, array_agg(val))
from test
group by id
)
'''
psycopg2 returns the 'result' as a basic string, while
in asyncpg and py-postgresql I have structured data
(roughly 'List[Tuple[int, List[str]]]').
I tried the same in psycopg3 and it is little bit better, but
not entirely: it shows the outer list, the tuples inside it,
but the innermost list is still represented as a basic string:
'{one,"one more"}'.
Is it something you are still working on? Any workarounds?
Vladimir
On Tue, 10 Nov 2020 at 21:59, Vladimir Ryabtsev <greatvovan@gmail.com> wrote: > psycopg2 returns the 'result' as a basic string, while > in asyncpg and py-postgresql I have structured data > (roughly 'List[Tuple[int, List[str]]]'). > > I tried the same in psycopg3 and it is little bit better, but > not entirely: it shows the outer list, the tuples inside it, > but the innermost list is still represented as a basic string: > '{one,"one more"}'. > > Is it something you are still working on? Any workarounds? Yes: by obtaining data from the db in binary mode you can get information about deeply nested objects. psycopg2 works only in text mode, psycopg3 in both. In [1]: query = """ ...: with test as ( ...: select 1 as id, 'one' val ...: union all ...: select 1, 'one more' ...: union all ...: select 2, 'two' ...: ) ...: select array( ...: select (id, array_agg(val)) ...: from test ...: group by id ...: )""" In [2]: import psycopg3 In [3]: from psycopg3.pq import Format In [4]: cnn = psycopg3.connect("") In [5]: cnn.cursor().execute(query).fetchone()[0] Out[5]: [('1', '{one,"one more"}'), ('2', '{two}')] In [6]: cnn.cursor(format=Format.BINARY).execute(query).fetchone()[0] Out[6]: [(1, ['one', 'one more']), (2, ['two'])] Binary loading/dumping is not supported yet for all the data types, but the plan is to cover all the builtins. Still not sure about the interface to request text/binary results, or whether binary shouldn't be the default as opposed to text. There is still ground to cover, but we are getting there. -- Daniele
Awesome, thanks.
It would be great if it worked out of the box, as in other drivers.
Appreciate your efforts.
Vladimir
On Tue, 10 Nov 2020 at 16:24, Daniele Varrazzo <daniele.varrazzo@gmail.com> wrote:
On Tue, 10 Nov 2020 at 21:59, Vladimir Ryabtsev <greatvovan@gmail.com> wrote:
> psycopg2 returns the 'result' as a basic string, while
> in asyncpg and py-postgresql I have structured data
> (roughly 'List[Tuple[int, List[str]]]').
>
> I tried the same in psycopg3 and it is little bit better, but
> not entirely: it shows the outer list, the tuples inside it,
> but the innermost list is still represented as a basic string:
> '{one,"one more"}'.
>
> Is it something you are still working on? Any workarounds?
Yes: by obtaining data from the db in binary mode you can get
information about deeply nested objects. psycopg2 works only in text
mode, psycopg3 in both.
In [1]: query = """
...: with test as (
...: select 1 as id, 'one' val
...: union all
...: select 1, 'one more'
...: union all
...: select 2, 'two'
...: )
...: select array(
...: select (id, array_agg(val))
...: from test
...: group by id
...: )"""
In [2]: import psycopg3
In [3]: from psycopg3.pq import Format
In [4]: cnn = psycopg3.connect("")
In [5]: cnn.cursor().execute(query).fetchone()[0]
Out[5]: [('1', '{one,"one more"}'), ('2', '{two}')]
In [6]: cnn.cursor(format=Format.BINARY).execute(query).fetchone()[0]
Out[6]: [(1, ['one', 'one more']), (2, ['two'])]
Binary loading/dumping is not supported yet for all the data types,
but the plan is to cover all the builtins. Still not sure about the
interface to request text/binary results, or whether binary shouldn't
be the default as opposed to text. There is still ground to cover, but
we are getting there.
-- Daniele
On Wed, 11 Nov 2020 at 00:36, Vladimir Ryabtsev <greatvovan@gmail.com> wrote: > > Awesome, thanks. > > It would be great if it worked out of the box, as in other drivers. I do agree. I have much more experience with the text format, and in my understanding not all the types support binary I/O. But my understanding could be wrong. After covering the binary format for the missing types (especially numeric and date/time objects) we can probably do more testing and check if defaulting to the binary format doesn't have unexpected consequences. Cheers! -- Daniele