Compressed CSV/JSON support (#3642)

* Compression text support

* Fix the path joining issue on Windows test

* Debug code for Windows CI

* Utilize `std::path::Path`, instead of `url::Url`

Author: Licht-T

datafusion-examples/examples/parquet_sql_multiple_files.rs

@@ -15,9 +15,8 @@
 // specific language governing permissions and limitations
 // under the License.
 
-use datafusion::datasource::file_format::parquet::{
-    ParquetFormat, DEFAULT_PARQUET_EXTENSION,
-};
+use datafusion::datasource::file_format::file_type::{FileType, GetExt};
+use datafusion::datasource::file_format::parquet::ParquetFormat;
 use datafusion::datasource::listing::ListingOptions;
 use datafusion::error::Result;
 use datafusion::prelude::*;
@@ -35,7 +34,7 @@ async fn main() -> Result<()> {
     // Configure listing options
     let file_format = ParquetFormat::default().with_enable_pruning(true);
     let listing_options = ListingOptions {
-        file_extension: DEFAULT_PARQUET_EXTENSION.to_owned(),
+        file_extension: FileType::PARQUET.get_ext(),
         format: Arc::new(file_format),
         table_partition_cols: vec![],
         collect_stat: true,

datafusion/common/src/error.rs

@@ -318,6 +318,12 @@ impl Display for DataFusionError {
 
 impl error::Error for DataFusionError {}
 
+impl From<DataFusionError> for io::Error {
+    fn from(e: DataFusionError) -> Self {
+        io::Error::new(io::ErrorKind::Other, e)
+    }
+}
+
 #[cfg(test)]
 mod test {
     use crate::error::DataFusionError;

datafusion/core/Cargo.toml

@@ -57,8 +57,10 @@ unicode_expressions = ["datafusion-physical-expr/regex_expressions", "datafusion
 ahash = { version = "0.8", default-features = false, features = ["runtime-rng"] }
 apache-avro = { version = "0.14", optional = true }
 arrow = { version = "24.0.0", features = ["prettyprint"] }
+async-compression = { version = "0.3.14", features = ["bzip2", "gzip", "futures-io", "tokio"] }
 async-trait = "0.1.41"
 bytes = "1.1"
+bzip2 = "0.4.3"
 chrono = { version = "0.4", default-features = false }
 datafusion-common = { path = "../common", version = "13.0.0", features = ["parquet", "object_store"] }
 datafusion-expr = { path = "../expr", version = "13.0.0" }
@@ -67,6 +69,7 @@ datafusion-optimizer = { path = "../optimizer", version = "13.0.0" }
 datafusion-physical-expr = { path = "../physical-expr", version = "13.0.0" }
 datafusion-row = { path = "../row", version = "13.0.0" }
 datafusion-sql = { path = "../sql", version = "13.0.0" }
+flate2 = "1.0.24"
 futures = "0.3"
 glob = "0.3.0"
 hashbrown = { version = "0.12", features = ["raw"] }
@@ -90,6 +93,7 @@ sqlparser = "0.25"
 tempfile = "3"
 tokio = { version = "1.0", features = ["macros", "rt", "rt-multi-thread", "sync", "fs", "parking_lot"] }
 tokio-stream = "0.1"
+tokio-util = { version = "0.7.4", features = ["io"] }
 url = "2.2"
 uuid = { version = "1.0", features = ["v4"] }
 
@@ -102,6 +106,7 @@ ctor = "0.1.22"
 doc-comment = "0.3"
 env_logger = "0.9"
 fuzz-utils = { path = "fuzz-utils" }
+rstest = "0.15.0"
 
 [[bench]]
 harness = false

datafusion/core/src/datasource/file_format/csv.rs

@@ -18,16 +18,21 @@
 //! CSV format abstractions
 
 use std::any::Any;
+
 use std::sync::Arc;
 
 use arrow::datatypes::Schema;
 use arrow::{self, datatypes::SchemaRef};
 use async_trait::async_trait;
+use bytes::Buf;
+
 use datafusion_common::DataFusionError;
+
 use futures::TryFutureExt;
 use object_store::{ObjectMeta, ObjectStore};
 
 use super::FileFormat;
+use crate::datasource::file_format::file_type::FileCompressionType;
 use crate::datasource::file_format::DEFAULT_SCHEMA_INFER_MAX_RECORD;
 use crate::error::Result;
 use crate::logical_plan::Expr;
@@ -43,6 +48,7 @@ pub struct CsvFormat {
     has_header: bool,
     delimiter: u8,
     schema_infer_max_rec: Option<usize>,
+    file_compression_type: FileCompressionType,
 }
 
 impl Default for CsvFormat {
@@ -51,6 +57,7 @@ impl Default for CsvFormat {
             schema_infer_max_rec: Some(DEFAULT_SCHEMA_INFER_MAX_RECORD),
             has_header: true,
             delimiter: b',',
+            file_compression_type: FileCompressionType::UNCOMPRESSED,
         }
     }
 }
@@ -82,6 +89,16 @@ impl CsvFormat {
         self
     }
 
+    /// Set a `FileCompressionType` of CSV
+    /// - defaults to `FileCompressionType::UNCOMPRESSED`
+    pub fn with_file_compression_type(
+        mut self,
+        file_compression_type: FileCompressionType,
+    ) -> Self {
+        self.file_compression_type = file_compression_type;
+        self
+    }
+
     /// The delimiter character.
     pub fn delimiter(&self) -> u8 {
         self.delimiter
@@ -110,8 +127,9 @@ impl FileFormat for CsvFormat {
                 .await
                 .map_err(|e| DataFusionError::External(Box::new(e)))?;
 
+            let decoder = self.file_compression_type.convert_read(data.reader());
             let (schema, records_read) = arrow::csv::reader::infer_reader_schema(
-                &mut data.as_ref(),
+                decoder,
                 self.delimiter,
                 Some(records_to_read),
                 self.has_header,
@@ -144,7 +162,12 @@ impl FileFormat for CsvFormat {
         conf: FileScanConfig,
         _filters: &[Expr],
     ) -> Result<Arc<dyn ExecutionPlan>> {
-        let exec = CsvExec::new(conf, self.has_header, self.delimiter);
+        let exec = CsvExec::new(
+            conf,
+            self.has_header,
+            self.delimiter,
+            self.file_compression_type.to_owned(),
+        );
         Ok(Arc::new(exec))
     }
 }