-
Notifications
You must be signed in to change notification settings - Fork 31
/
generate-input.py
34 lines (24 loc) · 1009 Bytes
/
generate-input.py
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
import pyarrow as pa
# Create a new field named "number" of type int64 that is not nullable.
number_field = pa.field('number', pa.int64(), nullable=False)
# Create a list of fields for pa.schema()
schema_fields = [number_field]
# Create a new schema from the fields.
schema = pa.schema(schema_fields)
# Construct some metadata to explain Fletchgen that it
# should allow the FPGA kernel to read from this schema.
metadata = {b'fletcher_mode': b'read',
b'fletcher_name': b'ExampleBatch'}
# Add the metadata to the schema
schema = schema.add_metadata(metadata)
# Create a list of PyArrow Arrays. Every Array can be seen
# as a 'Column' of the RecordBatch we will create.
data = [pa.array([1, -3, 3, -7])]
# Create a RecordBatch from the Arrays.
recordbatch = pa.RecordBatch.from_arrays(data, schema)
# Create an Arrow RecordBatchFileWriter.
writer = pa.RecordBatchFileWriter('recordbatch.rb', schema)
# Write the RecordBatch.
writer.write(recordbatch)
# Close the writer.
writer.close()