Thread: psycopg is the new psycopg3?

On 10/11/20, Daniele Varrazzo (daniele.varrazzo@gmail.com) wrote:
> Anyone against using "psycopg" as package name, and starting from 3 as
> version number?

Personally I think the "3" in "psycopg3" is clearer than simply
"psycopg" in relation to this being an evolution of the psycopg2
package.

Rory

On 11/10/20 9:24 AM, Rory Campbell-Lange wrote:
> On 10/11/20, Daniele Varrazzo (daniele.varrazzo@gmail.com) wrote:
>> Anyone against using "psycopg" as package name, and starting from 3 as
>> version number?
> 
> Personally I think the "3" in "psycopg3" is clearer than simply
> "psycopg" in relation to this being an evolution of the psycopg2
> package.

And for me, a lone 3 at the end of a package name means Python 3 version.

Maybe skip the 3 and go straight to 4?

--
~Ethan~

On 10/11/20 18:12, Daniele Varrazzo wrote:
> Hello,
> 
> I have received some concerned voices in regard to have a package called 
> "psycopg3". I guess many have been burned out by the Python 2 to 3 
> transition, and now it's not a happy pair of number to see next to each 
> other. Sorry, Fibonacci...
> 
> The rationale behind having the 2 in the package name was to allow the 
> coexistence between v1 and 2. But now that nobody uses v1 anymore, I 
> think the name can be considered free. I believe it even predates pypi 
> and the requirements.txt convention. Dark times...
> 
> Anyone against using "psycopg" as package name, and starting from 3 as 
> version number?

Seems fine to me.

federico

On Tue, Nov 10, 2020 at 6:12 PM Daniele Varrazzo
<daniele.varrazzo@gmail.com> wrote:
>
> Hello,
>
> I have received some concerned voices in regard to have a package called "psycopg3". I guess many have been burned
outby the Python 2 to 3 transition, and now it's not a happy pair of number to see next to each other. Sorry,
Fibonacci...
>
> The rationale behind having the 2 in the package name was to allow the coexistence between v1 and 2. But now that
nobodyuses v1 anymore, I think the name can be considered free. I believe it even predates pypi and the
requirements.txtconvention. Dark times... 
>
> Anyone against using "psycopg" as package name, and starting from 3 as version number?

This is a not entirely unsimilar case to what pgAdmin4 is going
through right now (they started with pgadmin4 version 1.0, which then
led to a lot of confusion for people).

Thus, regardless of if you call it psycopg or psycopg3, please make
sure you start with version 3 :)

--
 Magnus Hagander
 Me: https://www.hagander.net/
 Work: https://www.redpill-linpro.com/

On 11/10/20 9:12 AM, Daniele Varrazzo wrote:
> Hello,
> 
> I have received some concerned voices in regard to have a package called 
> "psycopg3". I guess many have been burned out by the Python 2 to 3 
> transition, and now it's not a happy pair of number to see next to each 
> other. Sorry, Fibonacci...
> 
> The rationale behind having the 2 in the package name was to allow the 
> coexistence between v1 and 2. But now that nobody uses v1 anymore, I 
> think the name can be considered free. I believe it even predates pypi 
> and the requirements.txt convention. Dark times...
> 
> Anyone against using "psycopg" as package name, and starting from 3 as 
> version number?

Yes.

1) "psycopg" is widely used as an alias for psycopg2, so that will cause 
confusion.

2) I see a lot of explaining why the order of versions is psycopg2, psycopg.

3) People don't seem to be confused that you can use psycopg2 with both 
Python 2 and 3.

> 
> Cheers,
> 
> -- Daniele


-- 
Adrian Klaver
adrian.klaver@aklaver.com

On 11/11/20 09:44, Matthieu Rigal wrote:
> I do also second that opinion. Also, the problem of "3" being connotated 
> in Python is fading out. Nobody should be scared by a "3" anymore...
> 
> My 3.9 cents
> 
> Le mer. 11 nov. 2020 à 09:30, Oleksandr Shulgin 
> <oleksandr.shulgin@zalando.de <mailto:oleksandr.shulgin@zalando.de>> a 
> écrit :
> 
>     On Tue, Nov 10, 2020 at 7:27 PM Vladimir Ryabtsev
>     <greatvovan@gmail.com <mailto:greatvovan@gmail.com>> wrote:
> 
>         As long as psycopg2 is not disappearing anywhere, psycopg3 seems
>         better.
>         New users will interpret psycopg2 as "more advanced psycopg".
> 
> 
>     I second that opinion.  Also, a suggestion to skip "3" and go with
>     "4" doesn't seem to help against the Python version confusion
>     concern, assuming there is Python4 in some (distant?) future.

Reading all the messages I have second toughts. If psycopg2 is here to 
stay, i.e., if it will not be completely replaced by "psycopg3" (and by 
completely I mean shutting down everything about it) then we will have 
the following situation:

psycopg2 version 2.x.y
psycopg  version 3.w.z

that at first sight is a bit confusing, isn't it?

federico

On Wed, 11 Nov 2020 at 08:47, Federico Di Gregorio <fog@dndg.it> wrote:

> Reading all the messages I have second toughts. If psycopg2 is here to
> stay, i.e., if it will not be completely replaced by "psycopg3" (and by
> completely I mean shutting down everything about it) then we will have
> the following situation:
>
> psycopg2 version 2.x.y
> psycopg  version 3.w.z
>
> that at first sight is a bit confusing, isn't it?

I have been settled with psycopg3 as the package name for a bit. Then,
a few days ago, releasing psycopg 2.9, I got to see some problems. The
main one is that, in order to respect semver, we should accept
introducing breaking changes only at the change of the main version.
People have been very confused to see breaking changes, although they
were minor, from 2.8 to 2.9.

Semver is much more an accepted, and expected, version number
organisation than having the major number in the package name. I can
expect to see psycopg 4, psycopg 5 etc. as we need to introduce
breaking changes. So I think, although going from psycopg2 v2.x to
psycopg v3.x might be confusing, the need to pin to the minor version
instead of the major is probably more so, and would come to bite us
much more often.

"import psycopg" is ready to merge
(https://github.com/psycopg/psycopg3/commit/7e526af8aca1c31b32a3ad55a0baf0de477c961c)

-- Daniele

On 6/26/21 4:48 AM, Daniele Varrazzo wrote:
> On Wed, 11 Nov 2020 at 08:47, Federico Di Gregorio <fog@dndg.it> wrote:
> 
>> Reading all the messages I have second toughts. If psycopg2 is here to
>> stay, i.e., if it will not be completely replaced by "psycopg3" (and by
>> completely I mean shutting down everything about it) then we will have
>> the following situation:
>>
>> psycopg2 version 2.x.y
>> psycopg  version 3.w.z
>>
>> that at first sight is a bit confusing, isn't it?
> 
> I have been settled with psycopg3 as the package name for a bit. Then,
> a few days ago, releasing psycopg 2.9, I got to see some problems. The
> main one is that, in order to respect semver, we should accept
> introducing breaking changes only at the change of the main version.
> People have been very confused to see breaking changes, although they
> were minor, from 2.8 to 2.9.
> 
> Semver is much more an accepted, and expected, version number
> organisation than having the major number in the package name. I can
> expect to see psycopg 4, psycopg 5 etc. as we need to introduce
> breaking changes. So I think, although going from psycopg2 v2.x to
> psycopg v3.x might be confusing, the need to pin to the minor version
> instead of the major is probably more so, and would come to bite us
> much more often.

Somewhere you lost me in the above.

What exactly is the proposed package naming and versioning going to be 
going forward?

> 
> "import psycopg" is ready to merge
> (https://github.com/psycopg/psycopg3/commit/7e526af8aca1c31b32a3ad55a0baf0de477c961c)
> 
> -- Daniele
> 
> 


-- 
Adrian Klaver
adrian.klaver@aklaver.com

Daniele Varrazzo a écrit :
> On Wed, 11 Nov 2020 at 08:47, Federico Di Gregorio <fog@dndg.it> wrote:
> 
>> Reading all the messages I have second toughts. If psycopg2 is here to
>> stay, i.e., if it will not be completely replaced by "psycopg3" (and by
>> completely I mean shutting down everything about it) then we will have
>> the following situation:
>>
>> psycopg2 version 2.x.y
>> psycopg  version 3.w.z
>>
>> that at first sight is a bit confusing, isn't it?
> 
> I have been settled with psycopg3 as the package name for a bit. Then,
> a few days ago, releasing psycopg 2.9, I got to see some problems. The
> main one is that, in order to respect semver, we should accept
> introducing breaking changes only at the change of the main version.
> People have been very confused to see breaking changes, although they
> were minor, from 2.8 to 2.9.
> 
> Semver is much more an accepted, and expected, version number
> organisation than having the major number in the package name. I can
> expect to see psycopg 4, psycopg 5 etc. as we need to introduce
> breaking changes. So I think, although going from psycopg2 v2.x to
> psycopg v3.x might be confusing, the need to pin to the minor version
> instead of the major is probably more so, and would come to bite us
> much more often.

That's a good point; so I agree moving to psycopg makes sense.

> "import psycopg" is ready to merge
> (https://github.com/psycopg/psycopg3/commit/7e526af8aca1c31b32a3ad55a0baf0de477c961c)

I see here that the repository (URL) is also renamed, which looks 
reasonable as well.

Am Mon, Jun 28, 2021 at 03:05:46PM +0200 schrieb Denis Laxalde:

> >>psycopg2 version 2.x.y
> >>psycopg  version 3.w.z
>
> That's a good point; so I agree moving to psycopg makes sense.

+1

Karsten
--
GPG  40BE 5B0E C98E 1713 AFA6  5BC0 3BEA AC80 7D4F C89B

Hi,

On Mon, 2021-06-28 at 15:05 +0200, Denis Laxalde wrote:
> That's a good point; so I agree moving to psycopg makes sense.

You folks also should talk to the RPM and deb packagers before making
the final decision ;)

RPM packager speaking: This change does not seem to hurt us.

Regards,

--
Devrim Gündüz
Open Source Solution Architect, Red Hat Certified Engineer
Twitter: @DevrimGunduz , @DevrimGunduzTR

Hi Devrim,

On Mon, 28 Jun 2021 at 21:49, Devrim Gündüz <devrim@gunduz.org> wrote:

> You folks also should talk to the RPM and deb packagers before making
> the final decision ;)
>
> RPM packager speaking: This change does not seem to hurt us.

Thank you, good call.

Package-wise, is there anything else we might do upstream to help you
there? I will ask the deb maintainers too.

-- Daniele

Devrim Gündüz a écrit :
> On Mon, 2021-06-28 at 15:05 +0200, Denis Laxalde wrote:
>> That's a good point; so I agree moving to psycopg makes sense.
> 
> You folks also should talk to the RPM and deb packagers before making
> the final decision ;)
> 
> RPM packager speaking: This change does not seem to hurt us.

There's no python3-psycopg package in the Debian archive so it probably 
won't hurt either.