Skip to content

Support writing IntervalMonthDayNanoArray to parquet via Arrow Writer#5868

Closed
xinlifoobar wants to merge 3 commits intoapache:masterfrom
xinlifoobar:dev/xinli/interval_monthdaynano
Closed

Support writing IntervalMonthDayNanoArray to parquet via Arrow Writer#5868
xinlifoobar wants to merge 3 commits intoapache:masterfrom
xinlifoobar:dev/xinli/interval_monthdaynano

Conversation

@xinlifoobar
Copy link
Contributor

Which issue does this PR close?

Closes #5849 and Related #5847

Rationale for this change

What changes are included in this PR?

Are there any user-facing changes?

@github-actions github-actions bot added the parquet Changes to the parquet crate label Jun 11, 2024
@xinlifoobar
Copy link
Contributor Author

This PR is super weird due to:

  1. From a spec perspective, it defines the interval data type as 12 bytes which could not accommodate the IntervalMonthDayNanoArray(16 bytes).
  2. From the arrow-rs implementations, the interval type is not defined as a logic type.

pub enum LogicalType {
String,
Map,
List,
Enum,
Decimal {
scale: i32,
precision: i32,
},
Date,
Time {
is_adjusted_to_u_t_c: bool,
unit: TimeUnit,
},
Timestamp {
is_adjusted_to_u_t_c: bool,
unit: TimeUnit,
},
Integer {
bit_width: i8,
is_signed: bool,
},
Unknown,
Json,
Bson,
Uuid,
Float16,
}

It would be more natural if we set both logic type and converted type at

pub struct PrimitiveTypeBuilder<'a> {
name: &'a str,
repetition: Repetition,
physical_type: PhysicalType,
converted_type: ConvertedType,
logical_type: Option<LogicalType>,
length: i32,
precision: i32,
scale: i32,
id: Option<i32>,
}

Now it is (None, ConvertedType:Interval).

@tustvold
Copy link
Contributor

#5849 (comment)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

parquet Changes to the parquet crate

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Support writing IntervalMonthDayNanoArray to parquet via Arrow Writer

2 participants