Use 'Numeric' instead of 'Float' as BQ column type when field type in the json schema is number #161
Conversation
| import com.snowplowanalytics.iglu.schemaddl.jsonschema.properties.NumberProperty | ||
| import com.snowplowanalytics.iglu.schemaddl.jsonschema.Schema | ||
|
|
||
| object decimals { |
There was a problem hiding this comment.
Can this be private to schemaddl?
| @@ -0,0 +1,28 @@ | |||
| package com.snowplowanalytics.iglu.schemaddl.jsonschema.suggestion | |||
|
|
|||
| object baseTypes { | |||
There was a problem hiding this comment.
Can this be private to schemaddl?
| @@ -42,27 +42,27 @@ object Field { | |||
| case Some(types) if types.possiblyWithNull(CommonProperties.Type.Object) => | |||
| val subfields = topSchema.properties.map(_.value).getOrElse(Map.empty) | |||
| if (subfields.isEmpty) { | |||
| Suggestion.finalSuggestion(topSchema, required)(name) | |||
| Suggestion.finalSuggestion(topSchema, required || types.nullable)(name) | |||
There was a problem hiding this comment.
As discussed in #154 I'm fairly sure this should be required && !types.nullable. To settle this disagreement I think we need a unit test that only passes with one version of the code.
| object baseTypes { | ||
| sealed trait BaseType extends Product with Serializable | ||
|
|
||
| object BaseType { |
There was a problem hiding this comment.
Maybe call it BaseNumericType? Or NumericType?
| else if ((scale == 0) && (max <= Long.MaxValue)) BaseType.Int64 | ||
| else BaseType.Decimal(max.precision - max.scale + scale, scale) | ||
|
|
||
| Some(NullableWrapper.fromBool(t, nullable)) |
There was a problem hiding this comment.
This is the only place where you ever call NullableWrapper.fromBool. I think I would prefer writing it out in full here, and then remove the fromBool function.
if (nullable) Some(NullableWrapper.NullableValue(t))
else Some(NullableWrapper.NotNullValue(t))|
|
||
| object NullableWrapper { | ||
|
|
||
| case class NullableValue(`type`: BaseType) extends NullableWrapper |
There was a problem hiding this comment.
In schema-ddl, we tend to use the word value only when there is a real data value (e.g. 42 or "hello"). If there is no data value, we tend to use the words like type or field.
voropaevp
force-pushed
the
fix/bigquery-suggestion-null
branch
from
f0997c2
to
b481d87
Compare
|
mixed up the branches on this one. Continued in #162. |
No description provided.