common_recordbatch/
recordbatch.rs

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
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
// Copyright 2023 Greptime Team
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

use std::collections::HashMap;
use std::slice;
use std::sync::Arc;

use datafusion::arrow::util::pretty::pretty_format_batches;
use datatypes::prelude::DataType;
use datatypes::schema::SchemaRef;
use datatypes::value::Value;
use datatypes::vectors::{Helper, VectorRef};
use serde::ser::{Error, SerializeStruct};
use serde::{Serialize, Serializer};
use snafu::{OptionExt, ResultExt};

use crate::error::{
    self, CastVectorSnafu, ColumnNotExistsSnafu, DataTypesSnafu, ProjectArrowRecordBatchSnafu,
    Result,
};
use crate::DfRecordBatch;

/// A two-dimensional batch of column-oriented data with a defined schema.
#[derive(Clone, Debug, PartialEq)]
pub struct RecordBatch {
    pub schema: SchemaRef,
    columns: Vec<VectorRef>,
    df_record_batch: DfRecordBatch,
}

impl RecordBatch {
    /// Create a new [`RecordBatch`] from `schema` and `columns`.
    pub fn new<I: IntoIterator<Item = VectorRef>>(
        schema: SchemaRef,
        columns: I,
    ) -> Result<RecordBatch> {
        let columns: Vec<_> = columns.into_iter().collect();
        let arrow_arrays = columns.iter().map(|v| v.to_arrow_array()).collect();

        let df_record_batch = DfRecordBatch::try_new(schema.arrow_schema().clone(), arrow_arrays)
            .context(error::NewDfRecordBatchSnafu)?;

        Ok(RecordBatch {
            schema,
            columns,
            df_record_batch,
        })
    }

    /// Create an empty [`RecordBatch`] from `schema`.
    pub fn new_empty(schema: SchemaRef) -> RecordBatch {
        let df_record_batch = DfRecordBatch::new_empty(schema.arrow_schema().clone());
        let columns = schema
            .column_schemas()
            .iter()
            .map(|col| col.data_type.create_mutable_vector(0).to_vector())
            .collect();
        RecordBatch {
            schema,
            columns,
            df_record_batch,
        }
    }

    pub fn try_project(&self, indices: &[usize]) -> Result<Self> {
        let schema = Arc::new(self.schema.try_project(indices).context(DataTypesSnafu)?);
        let mut columns = Vec::with_capacity(indices.len());
        for index in indices {
            columns.push(self.columns[*index].clone());
        }
        let df_record_batch = self.df_record_batch.project(indices).with_context(|_| {
            ProjectArrowRecordBatchSnafu {
                schema: self.schema.clone(),
                projection: indices.to_vec(),
            }
        })?;

        Ok(Self {
            schema,
            columns,
            df_record_batch,
        })
    }

    /// Create a new [`RecordBatch`] from `schema` and `df_record_batch`.
    ///
    /// This method doesn't check the schema.
    pub fn try_from_df_record_batch(
        schema: SchemaRef,
        df_record_batch: DfRecordBatch,
    ) -> Result<RecordBatch> {
        let columns = df_record_batch
            .columns()
            .iter()
            .map(|c| Helper::try_into_vector(c.clone()).context(error::DataTypesSnafu))
            .collect::<Result<Vec<_>>>()?;

        Ok(RecordBatch {
            schema,
            columns,
            df_record_batch,
        })
    }

    #[inline]
    pub fn df_record_batch(&self) -> &DfRecordBatch {
        &self.df_record_batch
    }

    #[inline]
    pub fn into_df_record_batch(self) -> DfRecordBatch {
        self.df_record_batch
    }

    #[inline]
    pub fn columns(&self) -> &[VectorRef] {
        &self.columns
    }

    #[inline]
    pub fn column(&self, idx: usize) -> &VectorRef {
        &self.columns[idx]
    }

    pub fn column_by_name(&self, name: &str) -> Option<&VectorRef> {
        let idx = self.schema.column_index_by_name(name)?;
        Some(&self.columns[idx])
    }

    #[inline]
    pub fn num_columns(&self) -> usize {
        self.columns.len()
    }

    #[inline]
    pub fn num_rows(&self) -> usize {
        self.df_record_batch.num_rows()
    }

    /// Create an iterator to traverse the data by row
    pub fn rows(&self) -> RecordBatchRowIterator<'_> {
        RecordBatchRowIterator::new(self)
    }

    pub fn column_vectors(
        &self,
        table_name: &str,
        table_schema: SchemaRef,
    ) -> Result<HashMap<String, VectorRef>> {
        let mut vectors = HashMap::with_capacity(self.num_columns());

        // column schemas in recordbatch must match its vectors, otherwise it's corrupted
        for (vector_schema, vector) in self.schema.column_schemas().iter().zip(self.columns.iter())
        {
            let column_name = &vector_schema.name;
            let column_schema =
                table_schema
                    .column_schema_by_name(column_name)
                    .context(ColumnNotExistsSnafu {
                        table_name,
                        column_name,
                    })?;
            let vector = if vector_schema.data_type != column_schema.data_type {
                vector
                    .cast(&column_schema.data_type)
                    .with_context(|_| CastVectorSnafu {
                        from_type: vector.data_type(),
                        to_type: column_schema.data_type.clone(),
                    })?
            } else {
                vector.clone()
            };

            let _ = vectors.insert(column_name.clone(), vector);
        }

        Ok(vectors)
    }

    /// Pretty display this record batch like a table
    pub fn pretty_print(&self) -> String {
        pretty_format_batches(slice::from_ref(&self.df_record_batch))
            .map(|t| t.to_string())
            .unwrap_or("failed to pretty display a record batch".to_string())
    }
}

impl Serialize for RecordBatch {
    fn serialize<S>(&self, serializer: S) -> std::result::Result<S::Ok, S::Error>
    where
        S: Serializer,
    {
        // TODO(yingwen): arrow and arrow2's schemas have different fields, so
        // it might be better to use our `RawSchema` as serialized field.
        let mut s = serializer.serialize_struct("record", 2)?;
        s.serialize_field("schema", &**self.schema.arrow_schema())?;

        let vec = self
            .columns
            .iter()
            .map(|c| c.serialize_to_json())
            .collect::<std::result::Result<Vec<_>, _>>()
            .map_err(S::Error::custom)?;

        s.serialize_field("columns", &vec)?;
        s.end()
    }
}

pub struct RecordBatchRowIterator<'a> {
    record_batch: &'a RecordBatch,
    rows: usize,
    columns: usize,
    row_cursor: usize,
}

impl<'a> RecordBatchRowIterator<'a> {
    fn new(record_batch: &'a RecordBatch) -> RecordBatchRowIterator<'a> {
        RecordBatchRowIterator {
            record_batch,
            rows: record_batch.df_record_batch.num_rows(),
            columns: record_batch.df_record_batch.num_columns(),
            row_cursor: 0,
        }
    }
}

impl Iterator for RecordBatchRowIterator<'_> {
    type Item = Vec<Value>;

    fn next(&mut self) -> Option<Self::Item> {
        if self.row_cursor == self.rows {
            None
        } else {
            let mut row = Vec::with_capacity(self.columns);

            for col in 0..self.columns {
                let column = self.record_batch.column(col);
                row.push(column.get(self.row_cursor));
            }

            self.row_cursor += 1;
            Some(row)
        }
    }
}

#[cfg(test)]
mod tests {
    use std::sync::Arc;

    use datatypes::arrow::datatypes::{DataType, Field, Schema as ArrowSchema};
    use datatypes::data_type::ConcreteDataType;
    use datatypes::schema::{ColumnSchema, Schema};
    use datatypes::vectors::{StringVector, UInt32Vector};

    use super::*;

    #[test]
    fn test_record_batch() {
        let arrow_schema = Arc::new(ArrowSchema::new(vec![
            Field::new("c1", DataType::UInt32, false),
            Field::new("c2", DataType::UInt32, false),
        ]));
        let schema = Arc::new(Schema::try_from(arrow_schema).unwrap());

        let c1 = Arc::new(UInt32Vector::from_slice([1, 2, 3]));
        let c2 = Arc::new(UInt32Vector::from_slice([4, 5, 6]));
        let columns: Vec<VectorRef> = vec![c1, c2];

        let batch = RecordBatch::new(schema.clone(), columns.clone()).unwrap();
        assert_eq!(3, batch.num_rows());
        assert_eq!(&columns, batch.columns());
        for (i, expect) in columns.iter().enumerate().take(batch.num_columns()) {
            let column = batch.column(i);
            assert_eq!(expect, column);
        }
        assert_eq!(schema, batch.schema);

        assert_eq!(columns[0], *batch.column_by_name("c1").unwrap());
        assert_eq!(columns[1], *batch.column_by_name("c2").unwrap());
        assert!(batch.column_by_name("c3").is_none());

        let converted =
            RecordBatch::try_from_df_record_batch(schema, batch.df_record_batch().clone()).unwrap();
        assert_eq!(batch, converted);
        assert_eq!(*batch.df_record_batch(), converted.into_df_record_batch());
    }

    #[test]
    pub fn test_serialize_recordbatch() {
        let column_schemas = vec![ColumnSchema::new(
            "number",
            ConcreteDataType::uint32_datatype(),
            false,
        )];
        let schema = Arc::new(Schema::try_new(column_schemas).unwrap());

        let numbers: Vec<u32> = (0..10).collect();
        let columns = vec![Arc::new(UInt32Vector::from_slice(numbers)) as VectorRef];
        let batch = RecordBatch::new(schema, columns).unwrap();

        let output = serde_json::to_string(&batch).unwrap();
        assert_eq!(
            r#"{"schema":{"fields":[{"name":"number","data_type":"UInt32","nullable":false,"dict_id":0,"dict_is_ordered":false,"metadata":{}}],"metadata":{"greptime:version":"0"}},"columns":[[0,1,2,3,4,5,6,7,8,9]]}"#,
            output
        );
    }

    #[test]
    fn test_record_batch_visitor() {
        let column_schemas = vec![
            ColumnSchema::new("numbers", ConcreteDataType::uint32_datatype(), false),
            ColumnSchema::new("strings", ConcreteDataType::string_datatype(), true),
        ];
        let schema = Arc::new(Schema::new(column_schemas));
        let columns: Vec<VectorRef> = vec![
            Arc::new(UInt32Vector::from_slice(vec![1, 2, 3, 4])),
            Arc::new(StringVector::from(vec![
                None,
                Some("hello"),
                Some("greptime"),
                None,
            ])),
        ];
        let recordbatch = RecordBatch::new(schema, columns).unwrap();

        let mut record_batch_iter = recordbatch.rows();
        assert_eq!(
            vec![Value::UInt32(1), Value::Null],
            record_batch_iter
                .next()
                .unwrap()
                .into_iter()
                .collect::<Vec<Value>>()
        );

        assert_eq!(
            vec![Value::UInt32(2), Value::String("hello".into())],
            record_batch_iter
                .next()
                .unwrap()
                .into_iter()
                .collect::<Vec<Value>>()
        );

        assert_eq!(
            vec![Value::UInt32(3), Value::String("greptime".into())],
            record_batch_iter
                .next()
                .unwrap()
                .into_iter()
                .collect::<Vec<Value>>()
        );

        assert_eq!(
            vec![Value::UInt32(4), Value::Null],
            record_batch_iter
                .next()
                .unwrap()
                .into_iter()
                .collect::<Vec<Value>>()
        );

        assert!(record_batch_iter.next().is_none());
    }
}