GitHub user hugokitano edited a discussion: Glue catalog updating
I'm able to write parquet and metadata/manifest files with the **Glue**
catalog, but I am not able to see the catalog update with snapshots, and
writing subsequently to the same table makes it clear the catalog always thinks
it's the first write. Confused about how I have to tell the catalog that I've
committed.
```
// instantiating catalog like so....
// let catalog_config = iceberg_catalog_glue::GlueCatalogConfig::builder()
// .warehouse(format!("s3://{}/{}", bucket_name, path_prefix).to_string())
// .build();
//
// let glue_catalog =
iceberg_catalog_glue::GlueCatalog::new(catalog_config).await
// .map_err(|e| iceberg::Error::new(
// iceberg::ErrorKind::Unexpected,
// format!("Failed to create Glue catalog: {}", e)
// ))?;
//
// IcebergCatalog::Glue(glue_catalog)
pub struct IcebergParquetWriteSession {
writer: Box<dyn IcebergWriter>,
table: Table,
tokio_runtime: Arc<tokio::runtime::Runtime>,
catalog: IcebergCatalog
}
pub fn commit_parquet_writes(
mut session: IcebergParquetWriteSession
) -> Result<Vec<IcebergDataFile>, iceberg::Error> {
session.tokio_runtime.block_on(async {
let data_files = session.writer.close().await?;
println!("Writer closed, got {} data files", data_files.len());
let tx = Transaction::new(&session.table);
println!("Transaction created for table {} at {}",
session.table.identifier().name(),
session.table.metadata().location());
println!("Table current snapshot before commit: {:?}",
session.table.metadata().current_snapshot_id());
let mut fast_append = tx.fast_append(None, vec![])?;
fast_append.add_data_files(data_files.clone())?;
println!("Added {} data files to transaction", data_files.len());
println!("Applying transaction...");
let updated_tx = fast_append.apply().await?;
println!("Transaction applied successfully");
// I tried adding the below to see if it works, but I get `Error:
FeatureUnsupported => Updating a table is not supported yet`
// println!("Committing transaction to catalog...");
// match &session.catalog {
// IcebergCatalog::Memory(mem_catalog) => {
// updated_tx.commit(mem_catalog).await?;
// },
// IcebergCatalog::Glue(glue_catalog) => {
// updated_tx.commit(glue_catalog).await?;
// }
// }
// println!("Transaction committed successfully");
let table_ident = session.table.identifier().clone();
let catalog_ref = match &session.catalog {
IcebergCatalog::Memory(mem_catalog) => {
let refreshed_table =
mem_catalog.load_table(&table_ident).await?;
println!("Table snapshot after commit: {:?}",
refreshed_table.metadata().current_snapshot_id());
},
IcebergCatalog::Glue(glue_catalog) => {
let refreshed_table =
glue_catalog.load_table(&table_ident).await?;
println!("Table snapshot after commit: {:?}",
refreshed_table.metadata().current_snapshot_id());
}
};
Ok(data_files)
})
}
```
I get the following output.
```
Writer closed, got 1 data files
Transaction created for table test_table_name at
s3://chalk-staging-data/hugo-test/iceberg_warehouse/test_glue_namespace/test_table_name
Table current snapshot before commit: None
Added 1 data files to transaction
Applying transaction...
Transaction applied successfully
Table snapshot after commit: None
```
In summary: with Glue/S3, I see all of the files I need, but the catalog
doesn't update, I have no snapshots, and subsequent writes don't have a parent
snapshot id i.e. think they are the first write.
GitHub link: https://github.com/apache/iceberg-rust/discussions/1168
----
This is an automatically sent email for issues@iceberg.apache.org.
To unsubscribe, please send an email to: issues-unsubscr...@iceberg.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@iceberg.apache.org
For additional commands, e-mail: issues-h...@iceberg.apache.org
