fix: copy into table collect files twice some times.

src/query/sql/src/planner/binder/copy_into_table.rs

@@ -223,6 +223,7 @@ impl Binder {
             write_mode: CopyIntoTableMode::Copy,
             query: None,
             enable_distributed: false,
+            files_collected: false,
         })
     }
 
@@ -400,6 +401,7 @@ impl Binder {
 
             enable_distributed: false,
             is_transform: false,
+            files_collected: true,
         };
 
         self.bind_copy_into_table_from_location(bind_context, plan)

src/query/sql/src/planner/plans/copy_into_table.rs

@@ -137,10 +137,16 @@ pub struct CopyIntoTablePlan {
     pub is_transform: bool,
 
     pub enable_distributed: bool,
+
+    pub files_collected: bool,
 }
 
 impl CopyIntoTablePlan {
     pub async fn collect_files(&mut self, ctx: &dyn TableContext) -> Result<()> {
+        if self.files_collected {
+            return Ok(());
+        }
+        self.files_collected = true;
         ctx.set_status_info("begin to list files");
         let start = Instant::now();
 

src/query/storages/parquet/src/parquet_rs/parquet_table/table.rs

@@ -49,6 +49,7 @@ use databend_common_storage::parquet_rs::read_metadata_async;
 use databend_common_storage::StageFileInfo;
 use databend_common_storage::StageFilesInfo;
 use databend_storages_common_table_meta::table::ChangeType;
+use log::info;
 use opendal::Operator;
 use parquet::file::metadata::ParquetMetaData;
 use parquet::schema::types::SchemaDescPtr;
@@ -174,7 +175,9 @@ impl ParquetRSTable {
         // Infer schema from the first parquet file.
         // Assume all parquet files have the same schema.
         // If not, throw error during reading.
-        let size = operator.stat(path).await?.content_length();
+        let stat = operator.stat(path).await?;
+        let size = stat.content_length();
+        info!("infer schema from file {}, with stat {:?}", path, stat);
         let first_meta = read_metadata_async(path, &operator, Some(size)).await?;
         let arrow_schema = infer_schema_with_extension(first_meta.file_metadata())?;
         let compression_ratio = get_compression_ratio(&first_meta);