I am working on spark project and came across interesting ( was known in
hive) convention spark use.
https://spark.apache.org/docs/2.3.0/sql-programming-guide.html#partition-discovery
in spark if I partition dataset. partition columns does not exists in
parquet schema and hence in final data file.
Iceberg stores all table columns in the underlying data files. It does not
store derived partition values in the data files. If you're partitioning by
date(ts), it won't store that date ordinal. If you're partitioning by
identity(date_col), it will store date_col.
When reading data, values from th
Hello folks,
I have been testing Iceberg reading with and without stats built into
Iceberg dataset manifest and found that there's a huge jump in network
traffic with the latter..
In my test I am comparing two Iceberg datasets, both written in Iceberg
format. One with and the other without stats
Thanks for bringing this up! My initial theory is that this table has a ton
of stats data that you have to read. That could happen in a couple of cases.
First, you might have large values in some columns. Parquet will suppress
its stats if values are larger than 4k and those are what Iceberg uses.
Thank you for reply and great explanation!
On Thu, Apr 18, 2019 at 8:54 AM Ryan Blue wrote:
> Iceberg stores all table columns in the underlying data files. It does not
> store derived partition values in the data files. If you're partitioning by
> date(ts), it won't store that date ordinal. If
Correction, partition count = 4308.
> Re: Changing the way we keep stats. Avro is a block splittable format and
is friendly with parallel compute frameworks like Spark.
Here I am trying to say that we don't need to change the format to columnar
right? The current format is already friendly for pa
Ah, my bad. I missed adding in the schema details .. Here are some details
on the dataset with stats :
Iceberg Schema Columns : 20
Spark Schema fields : 20
Snapshot Summary :{added-data-files=4308, added-records=11494037,
changed-partition-count=4308, total-records=11494037, total-data-files=43