AWS Glue：無法將 CSV gzip 解析為 Parquet

Question

我正在嘗試使用 AWS Glue ETL 工具將 CSV 文件解析為 Parquet。 我正在關注亞馬遜網站上的本教程。

由於我在 S3 存儲桶上有一堆 CSV 文件，因此我將它們的內容壓縮為 GZIP。 我正在使用以下 lambda 函數從 zip 下載數據，提取 CSV 文件並將其另存為我的 S3 存儲桶中的壓縮 CSV：

import json
import boto3
from botocore.vendored import requests
import zipfile
from gzip import GzipFile
from io import BytesIO

def lambda_handler(event, context):
    s3 = boto3.resource('s3')
    bucket_name = "my-bucket"
    file_url = 'http://dados.cvm.gov.br/dados/FI/DOC/CDA/DADOS/'
    file_name = "cda_fi_201801"

    req = requests.get(file_url + file_name+".zip", stream=True)
    data = req.raw.read()
    zf = zipfile.ZipFile(BytesIO(data))


    for fn in zf.namelist():

        bytes = zf.read(fn).decode("windows-1252")

        print ('File:', fn)
        print ('has',len(bytes),'bytes')

        # Choose folder name to put csv file
        parts = fn.split("_")
        folder = "PL"

        if(parts[2] == "BLC"):
            folder = "BLC_"+parts[3]

        # BytesIO to not save to disk
        gz_body = BytesIO()
        gz = GzipFile(None, 'wb', 9, gz_body)

        # Write csv bytes to gzip body
        gz.write(bytes.encode('utf8'))

        s3.Bucket(bucket_name).put_object(Key=folder + "/" + fn,
            ContentType="text/plain",
            ContentEncoding='gzip',
            Body=gz_body.getvalue())

        gz.close()

    return {
        'statusCode': 200,
    }

在 blc_1 文件夾上運行 AWS Glue Crawler，我得到了以下表屬性：

{
    "StorageDescriptor": {
        "cols": {
            "FieldSchema": [
                {
                    "name": "tp_fundo",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "cnpj_fundo",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "denom_social",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "dt_comptc",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "tp_aplic",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "tp_ativo",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "emissor_ligado",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "tp_negoc",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "qt_venda_negoc",
                    "type": "double",
                    "comment": ""
                },
                {
                    "name": "vl_venda_negoc",
                    "type": "double",
                    "comment": ""
                },
                {
                    "name": "qt_aquis_negoc",
                    "type": "double",
                    "comment": ""
                },
                {
                    "name": "vl_aquis_negoc",
                    "type": "double",
                    "comment": ""
                },
                {
                    "name": "qt_pos_final",
                    "type": "double",
                    "comment": ""
                },
                {
                    "name": "vl_merc_pos_final",
                    "type": "double",
                    "comment": ""
                },
                {
                    "name": "vl_custo_pos_final",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "dt_confid_aplic",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "tp_titpub",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "cd_isin",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "cd_selic",
                    "type": "bigint",
                    "comment": ""
                },
                {
                    "name": "dt_emissao",
                    "type": "string",
                    "comment": ""
                },
                {
                    "name": "dt_venc",
                    "type": "string",
                    "comment": ""
                }
            ]
        },
        "location": "s3://my-bucket/BLC_1/",
        "inputFormat": "org.apache.hadoop.mapred.TextInputFormat",
        "outputFormat": "org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat",
        "compressed": "true",
        "numBuckets": "-1",
        "SerDeInfo": {
            "name": "",
            "serializationLib": "org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe",
            "parameters": {
                "field.delim": ";"
            }
        },
        "bucketCols": [],
        "sortCols": [],
        "parameters": {
            "skip.header.line.count": "1",
            "sizeKey": "731056",
            "objectCount": "1",
            "UPDATED_BY_CRAWLER": "blc-1",
            "CrawlerSchemaSerializerVersion": "1.0",
            "recordCount": "1884",
            "averageRecordSize": "258",
            "CrawlerSchemaDeserializerVersion": "1.0",
            "compressionType": "gzip",
            "classification": "csv",
            "columnsOrdered": "true",
            "areColumnsQuoted": "false",
            "delimiter": ";",
            "typeOfData": "file"
        },
        "SkewedInfo": {},
        "storedAsSubDirectories": "false"
    },
    "parameters": {
        "skip.header.line.count": "1",
        "sizeKey": "731056",
        "objectCount": "1",
        "UPDATED_BY_CRAWLER": "blc-1",
        "CrawlerSchemaSerializerVersion": "1.0",
        "recordCount": "1884",
        "averageRecordSize": "258",
        "CrawlerSchemaDeserializerVersion": "1.0",
        "compressionType": "gzip",
        "classification": "csv",
        "columnsOrdered": "true",
        "areColumnsQuoted": "false",
        "delimiter": ";",
        "typeOfData": "file"
    }
}

之后，我按照教程嘗試使用 Glue 自動生成的以下腳本運行 ETL 作業：

import sys
from awsglue.transforms import *
from awsglue.utils import getResolvedOptions
from pyspark.context import SparkContext
from awsglue.context import GlueContext
from awsglue.job import Job

## @params: [JOB_NAME]
args = getResolvedOptions(sys.argv, ['JOB_NAME'])

sc = SparkContext()
glueContext = GlueContext(sc)
spark = glueContext.spark_session
job = Job(glueContext)
job.init(args['JOB_NAME'], args)
## @type: DataSource
## @args: [database = "cvm", table_name = "blc_1", transformation_ctx = "datasource0"]
## @return: datasource0
## @inputs: []
datasource0 = glueContext.create_dynamic_frame.from_catalog(database = "cvm", table_name = "blc_1", transformation_ctx = "datasource0")
## @type: ApplyMapping
## @args: [mapping = [("tp_fundo", "string", "tp_fundo", "string"), ("cnpj_fundo", "string", "cnpj_fundo", "string"), ("denom_social", "string", "denom_social", "string"), ("dt_comptc", "string", "dt_comptc", "string"), ("tp_aplic", "string", "tp_aplic", "string"), ("tp_ativo", "string", "tp_ativo", "string"), ("emissor_ligado", "string", "emissor_ligado", "string"), ("tp_negoc", "string", "tp_negoc", "string"), ("qt_venda_negoc", "double", "qt_venda_negoc", "double"), ("vl_venda_negoc", "double", "vl_venda_negoc", "double"), ("qt_aquis_negoc", "double", "qt_aquis_negoc", "double"), ("vl_aquis_negoc", "double", "vl_aquis_negoc", "double"), ("qt_pos_final", "double", "qt_pos_final", "double"), ("vl_merc_pos_final", "double", "vl_merc_pos_final", "double"), ("vl_custo_pos_final", "string", "vl_custo_pos_final", "string"), ("dt_confid_aplic", "string", "dt_confid_aplic", "string"), ("tp_titpub", "string", "tp_titpub", "string"), ("cd_isin", "string", "cd_isin", "string"), ("cd_selic", "long", "cd_selic", "long"), ("dt_emissao", "string", "dt_emissao", "string"), ("dt_venc", "string", "dt_venc", "string")], transformation_ctx = "applymapping1"]
## @return: applymapping1
## @inputs: [frame = datasource0]
applymapping1 = ApplyMapping.apply(frame = datasource0, mappings = [("tp_fundo", "string", "tp_fundo", "string"), ("cnpj_fundo", "string", "cnpj_fundo", "string"), ("denom_social", "string", "denom_social", "string"), ("dt_comptc", "string", "dt_comptc", "string"), ("tp_aplic", "string", "tp_aplic", "string"), ("tp_ativo", "string", "tp_ativo", "string"), ("emissor_ligado", "string", "emissor_ligado", "string"), ("tp_negoc", "string", "tp_negoc", "string"), ("qt_venda_negoc", "double", "qt_venda_negoc", "double"), ("vl_venda_negoc", "double", "vl_venda_negoc", "double"), ("qt_aquis_negoc", "double", "qt_aquis_negoc", "double"), ("vl_aquis_negoc", "double", "vl_aquis_negoc", "double"), ("qt_pos_final", "double", "qt_pos_final", "double"), ("vl_merc_pos_final", "double", "vl_merc_pos_final", "double"), ("vl_custo_pos_final", "string", "vl_custo_pos_final", "string"), ("dt_confid_aplic", "string", "dt_confid_aplic", "string"), ("tp_titpub", "string", "tp_titpub", "string"), ("cd_isin", "string", "cd_isin", "string"), ("cd_selic", "long", "cd_selic", "long"), ("dt_emissao", "string", "dt_emissao", "string"), ("dt_venc", "string", "dt_venc", "string")], transformation_ctx = "applymapping1")
## @type: ResolveChoice
## @args: [choice = "make_struct", transformation_ctx = "resolvechoice2"]
## @return: resolvechoice2
## @inputs: [frame = applymapping1]
resolvechoice2 = ResolveChoice.apply(frame = applymapping1, choice = "make_struct", transformation_ctx = "resolvechoice2")
## @type: DropNullFields
## @args: [transformation_ctx = "dropnullfields3"]
## @return: dropnullfields3
## @inputs: [frame = resolvechoice2]
dropnullfields3 = DropNullFields.apply(frame = resolvechoice2, transformation_ctx = "dropnullfields3")
## @type: DataSink
## @args: [connection_type = "s3", connection_options = {"path": "s3://my-bucket/blc_1"}, format = "parquet", transformation_ctx = "datasink4"]
## @return: datasink4
## @inputs: [frame = dropnullfields3]
datasink4 = glueContext.write_dynamic_frame.from_options(frame = dropnullfields3, connection_type = "s3", connection_options = {"path": "s3://my-bucket/blc_1"}, format = "parquet", transformation_ctx = "datasink4")
job.commit()

我收到以下錯誤：

19/03/27 19:10:07 警告 TaskSetManager：在 0.0 階段丟失任務 0.0（TID 0，ip-172-32-89-229.us-east-2.compute.internal，執行器 1）：com.amazonaws .services.glue.util.FatalException：無法解析文件：cda_fi_BLC_1_201801.csv

在 com.amazonaws.services.glue.readers.JacksonReader.hasNextFailSafe(JacksonReader.scala:94) 在 com.amazonaws.services.glue.readers.JacksonReader.hasNext(JacksonReader.scala:38) 在 com.amazonaws.services.glue .hadoop.TapeHadoopRecordReader.nextKeyValue(TapeHadoopRecordReader.scala:63) 在 org.apache.spark.rdd.NewHadoopRDD$$anon$1.hasNext(NewHadoopRDD.scala:207) 在 org.apache.spark.InterruptibleIterator.hasNext(InterruptibleIterator.scala :37) at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408) at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408) at org.apache.spark.shuffle。 sort.BypassMergeSortShuffleWriter.write(BypassMergeSortShuffleWriter.java:148) at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:96) at org.apache.spark.scheduler.ShuffleMapTask。

不幸的是，它沒有給出可能發生什么的線索。 當 CSV 內容未壓縮時，我能夠執行 ETL，所以我最好的猜測是我在 gzip 壓縮上做錯了什么，或者缺少一些配置。

如果你知道發生了什么，我很感激一些幫助。

謝謝！

Answer 1

這個 TapeHadoopRecordReader 非常敏感。 我建議你這樣做： df = spark.read.csv('s3://path')

然后您可以使用fromDF()將 df 轉換為動態幀