feat: Implement serialize/deserialize for datatypes #6
Conversation
| pub fn get_name(&self, name: &str) -> Option<&StructField> { | ||
| self.fields.iter().find(|field| field.name == name) | ||
| } | ||
| } |
There was a problem hiding this comment.
In Python and Java we also have convenient formatting of the struct. I think in rust this is the Display trait. A struct gets formatted as struct<string, int, long, double> based on the fields. Maybe also something to consider for rust
There was a problem hiding this comment.
I've included it in this PR. How is the formatting for maps and lists?
There was a problem hiding this comment.
Map: map<string, double>
List: list<string>
There was a problem hiding this comment.
Can you add this in a separate PR? This way we can get this one in.
| /// Time of day without date or timezone. | ||
| Time, | ||
| /// Timestamp without timezone | ||
| Timestamp, |
There was a problem hiding this comment.
Spark uses Timestamp to represent timestamp with timezone, TimestampNtz to represent timestamp without timezone
There was a problem hiding this comment.
I prefer to approach in this pr since it's same as iceberg's spec.
There was a problem hiding this comment.
I would also prefer to stick to the iceberg spec.
There was a problem hiding this comment.
I find the negation rather confusing, on top of that, we should encourage people to use timestamps without timezones =)
| /// Time of day without date or timezone. | ||
| Time, | ||
| /// Timestamp without timezone | ||
| Timestamp, |
There was a problem hiding this comment.
I prefer to approach in this pr since it's same as iceberg's spec.
src/spec/datatypes.rs
Outdated
| #[serde(rename = "struct", tag = "type")] | ||
| pub struct StructType { | ||
| /// Struct fields | ||
| pub fields: Vec<StructField>, |
There was a problem hiding this comment.
How about making this field private here? The get/get_name has a linear searching algorithm, it maybe inefficient for large structs with hundreds of fields. We may need a hashmap to index name/id to fields.
There was a problem hiding this comment.
I can create a lookup table to speed up the field access. Should I include it in this PR or should I create a new PR for it?
There was a problem hiding this comment.
Should I include it in this PR or should I create a new PR for it?
I think it's not a block of merging this PR. We can add a new issue to track this instead.
There was a problem hiding this comment.
I think it's not a block of merging this PR. We can add a new issue to track this instead.
I'm ok with putting lookup table in later PR, just suggest to make fields private for now.
There was a problem hiding this comment.
A lookup table for "id" is probably most important. Would you also create a lookup table for "name"?
There was a problem hiding this comment.
I think so, accessing by name is also quite frequently in query optimizer.
There was a problem hiding this comment.
In Python we have both, and we cache them. A Python example:
def test_schema_index_by_name_visitor(table_schema_nested: Schema) -> None:
"""Test index_by_name visitor function"""
table_schema_nested = schema.Schema(
NestedField(field_id=1, name="foo", field_type=StringType(), required=False),
NestedField(field_id=2, name="bar", field_type=IntegerType(), required=True),
NestedField(field_id=3, name="baz", field_type=BooleanType(), required=False),
NestedField(
field_id=4,
name="qux",
field_type=ListType(element_id=5, element_type=StringType(), element_required=True),
required=True,
),
NestedField(
field_id=6,
name="quux",
field_type=MapType(
key_id=7,
key_type=StringType(),
value_id=8,
value_type=MapType(key_id=9, key_type=StringType(), value_id=10, value_type=IntegerType(), value_required=True),
value_required=True,
),
required=True,
),
NestedField(
field_id=11,
name="location",
field_type=ListType(
element_id=12,
element_type=StructType(
NestedField(field_id=13, name="latitude", field_type=FloatType(), required=False),
NestedField(field_id=14, name="longitude", field_type=FloatType(), required=False),
),
element_required=True,
),
required=True,
),
NestedField(
field_id=15,
name="person",
field_type=StructType(
NestedField(field_id=16, name="name", field_type=StringType(), required=False),
NestedField(field_id=17, name="age", field_type=IntegerType(), required=True),
),
required=False,
),
schema_id=1,
identifier_field_ids=[1],
)
index = schema.index_by_name(table_schema_nested)
assert index == {
"foo": 1,
"bar": 2,
"baz": 3,
"qux": 4,
"qux.element": 5,
"quux": 6,
"quux.key": 7,
"quux.value": 8,
"quux.value.key": 9,
"quux.value.value": 10,
"location": 11,
"location.element": 12,
"location.element.latitude": 13,
"location.element.longitude": 14,
"location.latitude": 13,
"location.longitude": 14,
"person": 15,
"person.name": 16,
"person.age": 17,
}Let's do a separate PR!
Co-authored-by: Renjie Liu <liurenjie2008@gmail.com>
JanKaul
changed the title
|
Thank you @JanKaul for the PR! And @liurenjie1024, @ConeyLiu and @Xuanwo for the review! |
|
Thank you all for the review and the great input. |
This PR adds the support to serialize/deserialize the iceberg datatypes.