CODEWITHSUNDEEP
pythongoogle-cloud-storagegoogle-cloud-dataflowapache-beamapache-beam-io
user11118940
user11118940

Reputation:

how to write to GCS with a ParDo and a DoFn in apache beam

Using apache_beam.io.filesystems.FileSystems how to write to GCS with a ParDo and a DoFn ?? I am already getting output in csv format from a pardo, do i need to write another pardo to write it to gcs or can i directly import a module to write it directly to gcs ? please help

Upvotes: 0

Views: 3063

Answers (1)

Guillem Xercavins
Guillem Xercavins

Reputation: 7058

I have an example here where I write b64-encoded images to GCS using apache_beam.io.filesystems.FileSystems. The last step of the pipeline takes b64 as the PCollection containing two fields key_id and image and applies the ParDo:

b64 | 'Save images' >> beam.ParDo(WriteToSeparateFiles(known_args.output))

where known_args.output is a GCS base path and WriteToSeparateFiles is as follows:

class WriteToSeparateFiles(beam.DoFn):
    def __init__(self, outdir):
        self.outdir = outdir
    def process(self, element):
        writer = filesystems.FileSystems.create(self.outdir + element['key_id'] + '.png')
        writer.write(element['image'])
        writer.close()

With filesystems.FileSystems.create() I have control over the destination path. For the base path I use the parameter we passed to the function and I use the key_id of each element to generate meaningful file names. Finally, I append the .png extension as I am writing images.

I use writer.write(element['image']) to save the contents of the image field for each file and close the stream with writer.close().

Upvotes: 3

Related Questions