Open hvlawren opened 6 years ago
Adding to this, NULL values seem to be generally handled poorly in this module. Something else that should work, but does not:
{where: {booleanProp: {inq: [null, false]}}
will not return "unset" (= null
) rows, but only false
rows.
Or, even simpler: {where: {booleanProp: {neq: true}}
should turn into WHERE booleanprop IS NOT true
but actually turns into WHERE booleanprop != true
, which again excludes null
values.
Adding to this, NULL values seem to be generally handled poorly in this module. Something else that should work, but does not:
{where: {booleanProp: {inq: [null, false]}}
will not return "unset" (=null
) rows, but onlyfalse
rows. Or, even simpler:{where: {booleanProp: {neq: true}}
should turn intoWHERE booleanprop IS NOT true
but actually turns intoWHERE booleanprop != true
, which again excludesnull
values.
oh thanks I thought I was becoming crazy. Also ilike, like etc... doesn't work properly. Same if you give a or / and arrays.
I think a big part of this module needs to be corrected.
Any solution for the same issue, we need this feature critically to maintain soft delete with deleted_at timestamp column, can you please confirm is this issue from pg node.js driver or loopback Postgresql driver, we can offer some help with PR instead of making work around for our larger solution.
When the attribute is present in the JavaScript object with null, it is evident that the driver should send query with null
Here is the relevant code:
I suspect the problem is in PostgreSQL.prototype._buildWhere
, it seems to handle only the following operators:
and
or
inq
nin
between
regexp
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
not stale
Thank you all for participating in the discussion. I am afraid our bandwidth is limited and this issue is unlikely to make it to our milestone backlog. We rely on you, the community of users, to contribute the changes necessary to fix the issues described above.
I am proposing the following approach:
!=
but use IS DISTINCT FROM
or IS NOT
instead.This way we can preserve backwards compatibility for existing users and get more time to fix any edge cases that may be discovered when the new implementation is put into real-world use. Once we are confident about the new implementation, we can make the feature flag enabled by default and release a new semver-major version.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
Yes this is still relevant and happening, I think that either a note should be added to the docs about a workaround or some attention should be given to it!
I think that either a note should be added to the docs about a workaround or some attention should be given to it!
@zardilior could you please contribute that attention yourself and open a pull request with the necessary changes?
Hello @everyone, any news on this issue ?
same problem with me !!
Is this still an issue? Can get this to work still with postgres.
Description/Steps to reproduce
Expected result
Null entries are not equal to value so should be included
Additional information
I notice that the output sql from the connector is !=, whereas IS DISTINCT FROM would work for NULL values. As far as I can tell, there is no operator that will invoke IS DISTINCT FROM. If this won't be fixed/is considered not an issue, having this at least documented somewhere would be useful.