comments

Author: alamb

datafusion-examples/examples/parquet_index_advanced.rs

@@ -139,17 +139,19 @@ async fn main() -> Result<()> {
         .show()
         .await?;
     println!("Files pruned: {}\n", provider.index().last_num_pruned());
+    // the underlying parquet reader reads ranges from all three files
 
     // Run a query that uses the index to prune files.
     //
     // Using the predicate "value = 150", the IndexTable can skip reading file 1
-    // (max value 100) and file 3 (min value of 200)
+    // (max value 1000) and file 3 (min value of 2000)
     println!("** Select data, predicate `value = 1500`");
     ctx.sql("SELECT file_name, value FROM index_table WHERE value = 1500")
         .await?
         .show()
         .await?;
     println!("Files pruned: {}\n", provider.index().last_num_pruned());
+    // TODO why are there two calls to read bytes from the file?
 
     // Now, use a predicate on both the value and tag columns. The tag column
     // is stored in multiple pages and thus we can prune entire pages based on