1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302
//! # (De)serializing Arrows Streaming IPC format.
//!
//! Arrow Streaming IPC is a [binary format format](https://arrow.apache.org/docs/python/ipc.html).
//! It used for sending an arbitrary length sequence of record batches.
//! The format must be processed from start to end, and does not support random access.
//! It is different than IPC, if you can't deserialize a file with `IpcReader::new`, it's probably an IPC Stream File.
//!
//! ## Example
//!
//! ```rust
//! use polars_core::prelude::*;
//! use polars_io::prelude::*;
//! use std::io::Cursor;
//!
//!
//! let s0 = Series::new("days", &[0, 1, 2, 3, 4]);
//! let s1 = Series::new("temp", &[22.1, 19.9, 7., 2., 3.]);
//! let mut df = DataFrame::new(vec![s0, s1]).unwrap();
//!
//! // Create an in memory file handler.
//! // Vec<u8>: Read + Write
//! // Cursor<T>: Seek
//!
//! let mut buf: Cursor<Vec<u8>> = Cursor::new(Vec::new());
//!
//! // write to the in memory buffer
//! IpcStreamWriter::new(&mut buf).finish(&mut df).expect("ipc writer");
//!
//! // reset the buffers index after writing to the beginning of the buffer
//! buf.set_position(0);
//!
//! // read the buffer into a DataFrame
//! let df_read = IpcStreamReader::new(buf).finish().unwrap();
//! assert!(df.equals(&df_read));
//! ```
use std::io::{Read, Write};
use std::path::PathBuf;
use arrow::io::ipc::read::{StreamMetadata, StreamState};
use arrow::io::ipc::write::WriteOptions;
use arrow::io::ipc::{read, write};
use polars_core::prelude::*;
use crate::prelude::*;
use crate::shared::{finish_reader, ArrowReader, WriterFactory};
/// Read Arrows Stream IPC format into a DataFrame
///
/// # Example
/// ```
/// use polars_core::prelude::*;
/// use std::fs::File;
/// use polars_io::ipc::IpcStreamReader;
/// use polars_io::SerReader;
///
/// fn example() -> PolarsResult<DataFrame> {
/// let file = File::open("file.ipc").expect("file not found");
///
/// IpcStreamReader::new(file)
/// .finish()
/// }
/// ```
#[must_use]
pub struct IpcStreamReader<R> {
/// File or Stream object
reader: R,
/// Aggregates chunks afterwards to a single chunk.
rechunk: bool,
n_rows: Option<usize>,
projection: Option<Vec<usize>>,
columns: Option<Vec<String>>,
row_index: Option<RowIndex>,
metadata: Option<StreamMetadata>,
}
impl<R: Read> IpcStreamReader<R> {
/// Get schema of the Ipc Stream File
pub fn schema(&mut self) -> PolarsResult<Schema> {
Ok(Schema::from_iter(&self.metadata()?.schema.fields))
}
/// Get arrow schema of the Ipc Stream File, this is faster than creating a polars schema.
pub fn arrow_schema(&mut self) -> PolarsResult<ArrowSchema> {
Ok(self.metadata()?.schema)
}
/// Stop reading when `n` rows are read.
pub fn with_n_rows(mut self, num_rows: Option<usize>) -> Self {
self.n_rows = num_rows;
self
}
/// Columns to select/ project
pub fn with_columns(mut self, columns: Option<Vec<String>>) -> Self {
self.columns = columns;
self
}
/// Add a row index column.
pub fn with_row_index(mut self, row_index: Option<RowIndex>) -> Self {
self.row_index = row_index;
self
}
/// Set the reader's column projection. This counts from 0, meaning that
/// `vec![0, 4]` would select the 1st and 5th column.
pub fn with_projection(mut self, projection: Option<Vec<usize>>) -> Self {
self.projection = projection;
self
}
fn metadata(&mut self) -> PolarsResult<StreamMetadata> {
match &self.metadata {
None => {
let metadata = read::read_stream_metadata(&mut self.reader)?;
self.metadata = Option::from(metadata.clone());
Ok(metadata)
},
Some(md) => Ok(md.clone()),
}
}
}
impl<R> ArrowReader for read::StreamReader<R>
where
R: Read,
{
fn next_record_batch(&mut self) -> PolarsResult<Option<RecordBatch>> {
self.next().map_or(Ok(None), |v| match v {
Ok(stream_state) => match stream_state {
StreamState::Waiting => Ok(None),
StreamState::Some(chunk) => Ok(Some(chunk)),
},
Err(err) => Err(err),
})
}
}
impl<R> SerReader<R> for IpcStreamReader<R>
where
R: Read,
{
fn new(reader: R) -> Self {
IpcStreamReader {
reader,
rechunk: true,
n_rows: None,
columns: None,
projection: None,
row_index: None,
metadata: None,
}
}
fn set_rechunk(mut self, rechunk: bool) -> Self {
self.rechunk = rechunk;
self
}
fn finish(mut self) -> PolarsResult<DataFrame> {
let rechunk = self.rechunk;
let metadata = self.metadata()?;
let schema = &metadata.schema;
if let Some(columns) = self.columns {
let prj = columns_to_projection(&columns, schema)?;
self.projection = Some(prj);
}
let schema = if let Some(projection) = &self.projection {
apply_projection(&metadata.schema, projection)
} else {
metadata.schema.clone()
};
let ipc_reader =
read::StreamReader::new(&mut self.reader, metadata.clone(), self.projection);
finish_reader(
ipc_reader,
rechunk,
self.n_rows,
None,
&schema,
self.row_index,
)
}
}
/// Write a DataFrame to Arrow's Streaming IPC format
///
/// # Example
///
/// ```
/// use polars_core::prelude::*;
/// use polars_io::ipc::IpcStreamWriter;
/// use std::fs::File;
/// use polars_io::SerWriter;
///
/// fn example(df: &mut DataFrame) -> PolarsResult<()> {
/// let mut file = File::create("file.ipc").expect("could not create file");
///
/// IpcStreamWriter::new(&mut file)
/// .finish(df)
/// }
///
/// ```
#[must_use]
pub struct IpcStreamWriter<W> {
writer: W,
compression: Option<IpcCompression>,
pl_flavor: bool,
}
use arrow::record_batch::RecordBatch;
use crate::RowIndex;
impl<W> IpcStreamWriter<W> {
/// Set the compression used. Defaults to None.
pub fn with_compression(mut self, compression: Option<IpcCompression>) -> Self {
self.compression = compression;
self
}
pub fn with_pl_flavor(mut self, pl_flavor: bool) -> Self {
self.pl_flavor = pl_flavor;
self
}
}
impl<W> SerWriter<W> for IpcStreamWriter<W>
where
W: Write,
{
fn new(writer: W) -> Self {
IpcStreamWriter {
writer,
compression: None,
pl_flavor: false,
}
}
fn finish(&mut self, df: &mut DataFrame) -> PolarsResult<()> {
let mut ipc_stream_writer = write::StreamWriter::new(
&mut self.writer,
WriteOptions {
compression: self.compression.map(|c| c.into()),
},
);
ipc_stream_writer.start(&df.schema().to_arrow(self.pl_flavor), None)?;
let df = chunk_df_for_writing(df, 512 * 512)?;
let iter = df.iter_chunks(self.pl_flavor);
for batch in iter {
ipc_stream_writer.write(&batch, None)?
}
ipc_stream_writer.finish()?;
Ok(())
}
}
pub struct IpcStreamWriterOption {
compression: Option<IpcCompression>,
extension: PathBuf,
}
impl IpcStreamWriterOption {
pub fn new() -> Self {
Self {
compression: None,
extension: PathBuf::from(".ipc"),
}
}
/// Set the compression used. Defaults to None.
pub fn with_compression(mut self, compression: Option<IpcCompression>) -> Self {
self.compression = compression;
self
}
/// Set the extension. Defaults to ".ipc".
pub fn with_extension(mut self, extension: PathBuf) -> Self {
self.extension = extension;
self
}
}
impl Default for IpcStreamWriterOption {
fn default() -> Self {
Self::new()
}
}
impl WriterFactory for IpcStreamWriterOption {
fn create_writer<W: Write + 'static>(&self, writer: W) -> Box<dyn SerWriter<W>> {
Box::new(IpcStreamWriter::new(writer).with_compression(self.compression))
}
fn extension(&self) -> PathBuf {
self.extension.to_owned()
}
}