feat: Add ARRAY support #19
Conversation
google-cla
bot
added
the
cla: yes
|
@olavloite Can we rebase this PR so that it can be easier for reviewing? |
Certainly, it should now be rebased and without conflicts. |
| case sppb.TypeCode_NUMERIC: | ||
| var v spanner.NullNumeric | ||
| if err := col.Decode(&v); err != nil { | ||
| return err | ||
| } | ||
| dest[i] = v.Numeric | ||
| dest[i] = v |
There was a problem hiding this comment.
Yeah, I can see the confusion here. I've add a comment for why we are assigning v here in all cases (that is; also when v.Valid is false). The reason that we do so for NUMERIC and DATE data types is that the Go sql package does not have a native type mapping for these types. That means that if we were to return nil here, Go sql would not correctly understand what type it is. So instead we return a NullNumeric or NullDate in these cases.
The reason we don't do that for for example INT64 is that the Go sql package natively maps the Spanner INT64 data type to the Go int64 type, and so we should return nil in those cases.
| } else { | ||
| dest[i] = v.Date.In(time.UTC) // TODO(jbd): Add note about this. | ||
| } | ||
| dest[i] = v |
Should not be merged before #10 (or should be merged instead of #10, as they both contain the same changes).