GitHub user ryux1 closed the discussion with a comment: How to get Parquet's
SchemaDescriptor when using parquet ArrowWriter?
`ArrowWriter::finish()` (or `close()`) returns `ParquetMetaData`, so you can
get the descriptor from its file metadata directly. You do not need to
reconstruct it from the Thrift `schema` field.
```rust
let metadata = writer.finish()?;
let schema = metadata.file_metadata().schema_descr();
for row_group in metadata.row_groups() {
for (i, column) in row_group.columns().iter().enumerate() {
let descriptor = schema.column(i);
if let Some(statistics) = column.statistics() {
println!("{}: {statistics:?}", descriptor.path());
// Match on `statistics` here if you need typed min/max values.
}
}
}
```
The column index here is the Parquet physical leaf-column index, which is why
iterating `row_group.columns()` and looking up `schema.column(i)` is preferable
to indexing by top-level Arrow fields.
If you need metadata before finishing, `writer.flushed_row_groups()` exposes
only row groups that have already been flushed; each `RowGroupMetaData` also
has `schema_descr()`. For complete file metadata, use the value returned by
`finish()`/`close()`.
I checked this against Arrow Rust 57.0.0 (the release current when this
question was opened) with a small compiling writer example. Disclosure: I used
OpenAI Codex to help inspect the relevant APIs and validate the example.
GitHub link:
https://github.com/apache/arrow-rs/discussions/8705#discussioncomment-18340597
----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]
