閱讀分佈式製表符分隔的CSV

Question

受此啓發question，我編寫了一些代碼來存儲RDD（從Parquet文件中讀取），其中包含（photo_id，data）的Schema（成對），並由製表符分隔，以及只是作爲一個詳細基地64編碼，就像這樣：

def do_pipeline(itr): 
    ... 
    item_id = x.photo_id 

def toTabCSVLine(data): 
    return '\t'.join(str(d) for d in data) 

serialize_vec_b64pkl = lambda x: (x[0], base64.b64encode(cPickle.dumps(x[1]))) 

def format(data): 
    return toTabCSVLine(serialize_vec_b64pkl(data)) 

dataset = sqlContext.read.parquet('mydir') 
lines = dataset.map(format) 
lines.saveAsTextFile('outdir') 

所以，現在的關注點：如何讀取數據集和打印，例如它的反序列化的數據？

我正在使用Python 2.6.6。

---

我的企圖就在這裏，在這裏只是證實一切可以做到的，我寫了這個代碼：

deserialize_vec_b64pkl = lambda x: (x[0], cPickle.loads(base64.b64decode(x[1]))) 

base64_dataset = sc.textFile('outdir') 
collected_base64_dataset = base64_dataset.collect() 
print(deserialize_vec_b64pkl(collected_base64_dataset[0].split('\t'))) 

這就要求collect()，這對於測試是確定的，但在現實世界方案將難以...

---

編輯：

當我試圖zero323的建議：

foo = (base64_dataset.map(str.split).map(deserialize_vec_b64pkl)).collect() 

我得到這個錯誤，這歸結爲：

PythonRDD[2] at RDD at PythonRDD.scala:43 
16/08/04 18:32:30 WARN TaskSetManager: Lost task 4.0 in stage 0.0 (TID 4, gsta31695.tan.ygrid.yahoo.com): org.apache.spark.api.python.PythonException: Traceback (most recent call last): 
    File "/grid/0/tmp/yarn-local/usercache/gsamaras/appcache/application_1470212406507_56888/container_e04_1470212406507_56888_01_000009/pyspark.zip/pyspark/worker.py", line 98, in main 
    command = pickleSer._read_with_length(infile) 
    File "/grid/0/tmp/yarn-local/usercache/gsamaras/appcache/application_1470212406507_56888/container_e04_1470212406507_56888_01_000009/pyspark.zip/pyspark/serializers.py", line 164, in _read_with_length 
    return self.loads(obj) 
    File "/grid/0/tmp/yarn-local/usercache/gsamaras/appcache/application_1470212406507_56888/container_e04_1470212406507_56888_01_000009/pyspark.zip/pyspark/serializers.py", line 422, in loads 
    return pickle.loads(obj) 
UnpicklingError: NEWOBJ class argument has NULL tp_new 

    at org.apache.spark.api.python.PythonRunner$$anon$1.read(PythonRDD.scala:166) 
    at org.apache.spark.api.python.PythonRunner$$anon$1.<init>(PythonRDD.scala:207) 
    at org.apache.spark.api.python.PythonRunner.compute(PythonRDD.scala:125) 
    at org.apache.spark.api.python.PythonRDD.compute(PythonRDD.scala:70) 
    at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:306) 
    at org.apache.spark.rdd.RDD.iterator(RDD.scala:270) 
    at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66) 
    at org.apache.spark.scheduler.Task.run(Task.scala:89) 
    at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:227) 
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) 
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) 
    at java.lang.Thread.run(Thread.java:745) 

16/08/04 18:32:30 ERROR TaskSetManager: Task 12 in stage 0.0 failed 4 times; aborting job 
16/08/04 18:32:31 WARN TaskSetManager: Lost task 14.3 in stage 0.0 (TID 38, gsta31695.tan.ygrid.yahoo.com): TaskKilled (killed intentionally) 
16/08/04 18:32:31 WARN TaskSetManager: Lost task 13.3 in stage 0.0 (TID 39, gsta31695.tan.ygrid.yahoo.com): TaskKilled (killed intentionally) 
16/08/04 18:32:31 WARN TaskSetManager: Lost task 16.3 in stage 0.0 (TID 42, gsta31695.tan.ygrid.yahoo.com): TaskKilled (killed intentionally) 
--------------------------------------------------------------------------- 
Py4JJavaError        Traceback (most recent call last) 
/homes/gsamaras/code/read_and_print.py in <module>() 
    17  print(base64_dataset.map(str.split).map(deserialize_vec_b64pkl)) 
    18 
---> 19  foo = (base64_dataset.map(str.split).map(deserialize_vec_b64pkl)).collect() 
    20  print(foo) 

/home/gs/spark/current/python/lib/pyspark.zip/pyspark/rdd.py in collect(self) 
    769   """ 
    770   with SCCallSiteSync(self.context) as css: 
--> 771    port = self.ctx._jvm.PythonRDD.collectAndServe(self._jrdd.rdd()) 
    772   return list(_load_from_socket(port, self._jrdd_deserializer)) 
    773 

/home/gs/spark/current/python/lib/py4j-0.9-src.zip/py4j/java_gateway.py in __call__(self, *args) 
    811   answer = self.gateway_client.send_command(command) 
    812   return_value = get_return_value(
--> 813    answer, self.gateway_client, self.target_id, self.name) 
    814 
    815   for temp_arg in temp_args: 

/home/gs/spark/current/python/lib/py4j-0.9-src.zip/py4j/protocol.py in get_return_value(answer, gateway_client, target_id, name) 
    306     raise Py4JJavaError(
    307      "An error occurred while calling {0}{1}{2}.\n". 
--> 308      format(target_id, ".", name), value) 
    309    else: 
    310     raise Py4JError(

Py4JJavaError: An error occurred while calling z:org.apache.spark.api.python.PythonRDD.collectAndServe. 

Answer 1

讓我們嘗試一個簡單的例子。爲方便起見，我將使用便利的toolz庫，但在這裏並不是必需的。

import sys 
import base64 

if sys.version_info < (3,): 
    import cPickle as pickle 
else: 
    import pickle 


from toolz.functoolz import compose 

rdd = sc.parallelize([(1, {"foo": "bar"}), (2, {"bar": "foo"})]) 

現在，您的代碼現在不是完全可移植的。在Python 2中，base64.b64encode返回str，而在Python 3中返回bytes。讓我們表明：

• 的Python 2

  type(base64.b64encode(pickle.dumps({"foo": "bar"}))) 
  ## str 
  

• 的Python 3

  type(base64.b64encode(pickle.dumps({"foo": "bar"}))) 
  ## bytes 
  

所以讓我們添加到解碼管線：

# Equivalent to 
# def pickle_and_b64(x): 
#  return base64.b64encode(pickle.dumps(x)).decode("ascii") 

pickle_and_b64 = compose(
    lambda x: x.decode("ascii"), 
    base64.b64encode, 
    pickle.dumps 
) 

請注意，這不承擔任何特定形狀的數據。正因爲如此，我們將使用mapValues連載僅鍵：

serialized = rdd.mapValues(pickle_and_b64) 
serialized.first() 
## 1, u'KGRwMApTJ2ZvbycKcDEKUydiYXInCnAyCnMu') 

現在我們可以用簡單的格式遵循它並保存：

from tempfile import mkdtemp 
import os 

outdir = os.path.join(mkdtemp(), "foo") 

serialized.map(lambda x: "{0}\t{1}".format(*x)).saveAsTextFile(outdir) 

讀取該文件，我們逆轉這一過程：

# Equivalent to 
# def b64_and_unpickle(x): 
#  return pickle.loads(base64.b64decode(x)) 

b64_and_unpickle = compose(
    pickle.loads, 
    base64.b64decode 
) 

decoded = (sc.textFile(outdir) 
    .map(lambda x: x.split("\t")) # In Python 3 we could simply use str.split 
    .mapValues(b64_and_unpickle)) 

decoded.first() 
## (u'1', {'foo': 'bar'})