On Sunday, March 26, 2017 at 6:41:29 PM UTC-7, Andrew Burleson wrote:
>
> Would it be practical to add an alternative -- or perhaps this already
>>> works -- of just using strings? eg.
>>>
>>> foo.join('users.id' => 'id')
>>>
>>
>> In most cases in Sequel, ruby strings are treated as SQL strings, so this
>> wouldn't work. There are a few cases where an SQL string is not valid
>> where Sequel will treat a ruby string as an SQL identifier, but Sequel
>> quotes identifiers by default, so you can't use . inside a string for a
>> qualified identifier. You can do Sequel.lit('users.id') to drop down to
>> literal SQL, but that limits Sequel's introspection capabilities.
>>
>>
>>> That, or perhaps the symbolized equivalent ( `:"users.id"`) would solve
>>> pretty much every use of symbol splitting for me.
>>>
>>
>> That also doesn't work because Sequel quotes identifiers by default.
>>
>
> So then in the next major version of Sequel, what would your preferred
> manner of qualifying column names be? The symbol subscript approach, eg.
> :table_name[:column_name] ?
>
This requires modifying a core class (Symbol), and therefore cannot be
supported in core Sequel, it must remain an extension (symbol_aref). That
being said, symbol_aref is what I use in most of my applications.
> I think whatever method you prefer would be fine, but since queries that
> require qualifying the table name to avoid ambiguity are a relatively
> common case, I think it's nice to have a "standard" way to do this, and
> preferable for the library to have "one way" to do most things rather than
> may ways to do them.
>
The standard way is a bit long:
Sequel[:table_name][:column_name]
The other, shorter ways require extensions. So users have to decide
whether they want to modify a core class to allow shorter code.
Thus, the other thought I'd have for you is, since you're introducing
> deprecation warnings now, and since I presume your goal is to improve the
> stability and consistency of the library with major versions, is that you
> might want to just warn that the behavior is deprecated with 4.41 and
> remove it on 5.0, rather than have it survive as an "optional feature"
> through to 6.0 or later.
>
Removing the support completely would be nice, but I thought it may make
upgrading too difficult for applications that used symbol splitting
extensively (even though there exists a mostly automated conversion tool:
https://github.com/jeremyevans/sequel-unsplit). By just changing the
default, you could always turn it back on if you want to use it in your
application.
Thanks,
Jeremy
--
You received this message because you are subscribed to the Google Groups
"sequel-talk" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/sequel-talk.
For more options, visit https://groups.google.com/d/optout.
