"Mike Bayer" <[email protected]> writes:

> On Wed, Nov 30, 2022, at 10:30 AM, Lele Gaifax wrote:
>> On the former one, it would be nice if we could reduce the "distance"
>> between the new Range class and what popular drivers provide:
>> 
>> - "empty" vs "isempty": both asyncpg and psycopg have a "isempty"
>>   property, we could either rename the "empty" attribute or add an
>>   property alias
>> 
>> - bounds inclusiveness: both drivers expose that thru
>>   "lower_inc"/"upper_inc" properties, while in SA we must say
>>   `range.bounds[0] == '['`
>> 
>> - infinite bounds: likewise, they both provide "lower_inf"/"upper_inf"
>>   properties, whereas we must say `range.lower is None and not
>>   range.empty`
>> 
>> As you see, these are "syntactic sugar" differences, easily implemented
>> with some @properties.
>
> great, we can add those.

Ok, I'll try to propose a issue/PR in the following days.

>> 
>> What is more substantial, that required slightly more impacting changes
>> in a few cases, is that the SA Range does not have an exact notion of
>> its "type", i.e. there is no way to determine the data type of an empty
>> or infinite range.
>> ... 
>> So I wonder if it would be reasonable/doable/affordable if SA could grow
>> a "family" of classes, derived from current Range, one for each data
>> type, so that for example a DATARANGE column would be associated with a
>> DateRange instance.

> how about we add a "type" field to Range itself?    That would be an
> easier change and also remove the need for applications to deal with
> all kinds of different Range subclasses that mostly have identical
> behavior.

Yes, it could be done that way too, but from the developer perspective,
how is he going to create Range instances? Changing the constructor to
require a "type" argument seems uglier than using a specific class.

>
> type-specific Range methods could be called off of each AbstractRange
> type subclass, such as how the AbstractRange._resolve_for_literal()
> method works.

I will need to study that part.

>
> The Range object itself should also become a typing.Generic so that we
> could pass around the type as Range[int] or Range[datetime], etc.

Yes!

bye, lele.
-- 
nickname: Lele Gaifax | Quando vivrò di quello che ho pensato ieri
real: Emanuele Gaifas | comincerò ad aver paura di chi mi copia.
[email protected]  |                 -- Fortunato Depero, 1929.

-- 
SQLAlchemy - 
The Python SQL Toolkit and Object Relational Mapper

http://www.sqlalchemy.org/

To post example code, please provide an MCVE: Minimal, Complete, and Verifiable 
Example.  See  http://stackoverflow.com/help/mcve for a full description.
--- 
You received this message because you are subscribed to the Google Groups 
"sqlalchemy" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/sqlalchemy/878rjsfddq.fsf%40metapensiero.it.

Reply via email to