I think you should not have added separate collections/iterables as
separate values in the first place... all bets are off on how Solr behaves.

On Fri, Feb 6, 2026 at 6:08 AM Nicolò Rinaldi <[email protected]> wrote:

> Hi
>
> I’m here to ask for some clarification about a behaviour that I encountered
> while trying to index matrices with RemoveBlankFieldUpdateProcessorFactory
> in
> the update processor chain, expecting an error since this syntax isn't
> supported in Solr. Here's an example: if I try to index a multivalued field
> with values [["cat", "dog"], ["lemon", "orange"]], the field is modified to
> ["cat", "dog", "lemon", "orange"]—flattening the 2x2 "matrix" into a 4-item
> vector. This flattening seems to happen in the mutate function
> <
> https://github.com/SeaseLtd/solr/blob/c99af207c761ec34812ef1cc3054eb2804b7448b/solr/core/src/java/org/apache/solr/update/processor/FieldValueMutatingUpdateProcessor.java#L60
> >
> of FieldValueMutatingUpdateProcessor.
>
> This allows the field to be populated even though the syntax is incorrect.
> Is there a reason for this behaviour?
>
> Thanks,
>
> Nicolò
>

Reply via email to