Re: jsonb_set() strictness considered harmful to data - Mailing list pgsql-general

From David G. Johnston
Subject Re: jsonb_set() strictness considered harmful to data
Date
Msg-id CAKFQuwZ2w5D1yJ9vaexiR8kbqOyxbzKwo2hsbi74ruVfFiNe-g@mail.gmail.com
Whole thread Raw
In response to Re: jsonb_set() strictness considered harmful to data  (Christoph Moench-Tegeder <cmt@burggraben.net>)
Responses Re: jsonb_set() strictness considered harmful to data
List pgsql-general
On Fri, Oct 18, 2019 at 2:50 PM Christoph Moench-Tegeder <cmt@burggraben.net> wrote:
## Ariadne Conill (ariadne@dereferenced.org):

>    update users set info=jsonb_set(info, '{bar}', info->'foo');
>
> Typically, this works nicely, except for cases where evaluating
> info->'foo' results in an SQL null being returned.  When that happens,
> jsonb_set() returns an SQL null, which then results in data loss.[3]

So why don't you use the facilities of SQL to make sure to only
touch the rows which match the prerequisites?

  UPDATE users SET info = jsonb_set(info, '{bar}', info->'foo')
    WHERE info->'foo' IS NOT NULL;


There are many ways to add code to queries to make working with this function safer - though using them presupposes one remembers at the time of writing the query that there is danger and caveats in using this function.  I agree that we should have (and now) provided sane defined behavior when one of the inputs to the function is null instead blowing off the issue and defining the function as being strict.  Whether that is "ignore and return the original object" or "add the key with a json null scalar value" is debatable but either is considerably more useful than returning SQL NULL.

David J.

pgsql-general by date:

Previous
From: Edilmar Alves
Date:
Subject: Replication of Replication
Next
From: Ariadne Conill
Date:
Subject: Re: jsonb_set() strictness considered harmful to data