I am new to Python and spark.
We are using Azure Databrick and with help of PySpark code shown below for it.
data=spark.sql("SELECT 'Name' as name, 'Number' as number FROM Student")
print(data)
This solution will work for you.
from pyspark.sql.types import StructType,StructField, StringType, IntegerType
data2 = [("Finance",10),
("Marketing",20),
("Sales",30),
("IT",40)
]
schema = StructType([ \
StructField("Name",StringType(),True), \
StructField("number", IntegerType(), True) \
])
df = spark.createDataFrame(data=data2,schema=schema)
df1 = df.withColumn("Student",lit("Student")).select("Student",to_json(struct("Name","number")).alias("Data"))
display(df1)
according to the instructions in Colab I could get buffer & even take a pd.DataFrame from it (file is just example)...
# ... authentification
file_id = '1S1w0Z7g3bI1PGLPR49PW5VBRo7c_KYgU' # titanic
# loading data
import io
from googleapiclient.http import MediaIoBaseDownload
drive_service = build('drive', 'v3') # , credentials=creds
request = drive_service.files().get_media(fileId=file_id)
buf = io.BytesIO()
downloader = MediaIoBaseDownload(buf, request)
buf.seek(0)
import pandas as pd
df= pd.read_csv(buf);
print(df.head())
But have trouble with correct creation of dataFlow to Dataset - "buf" var is not working in =>
dataset = tf.data.experimental.make_csv_dataset(csv_file_path,
batch_size=100, num_epochs=1)
only "csv_file_path" as 1st arg. Is it possible in Colab to get IO from my GoogleDrive's csv-file into Dataset (used further in training)? And how to do it in a memory-efficient manner?..
P.S.
I understand that I perhaps can make file opened for all (in GoogleDrive) & get url to use the simple way:
#TRAIN_DATA_URL = "https://storage.googleapis.com/tf-datasets/titanic/train.csv"
TRAIN_DATA_URL = "https://drive.google.com/file/d/1S1w0Z7g3bI1PGLPR49PW5VBRo7c_KYgU/view?usp=sharing"
train_file_path = tf.keras.utils.get_file("train.csv", TRAIN_DATA_URL)
dataset = tf.data.experimental.make_csv_dataset(train_file_path, batch_size=100, num_epochs=1)
! but I DON'T need to share real file... How to save file confidential & get IO from it (in GoogleDrive) to tf.data.Dataset in Colab ? (preferably the shortest code - there will be much more code in real project tested in Colab)
drive.CreateFile HELPED (link) - as so as I understand that working in Colab - I am working in a separate environment (separate from my PC & I'net env)... So I tried (according link)
!pip install -U -q PyDrive
from pydrive.auth import GoogleAuth
from pydrive.drive import GoogleDrive
from google.colab import auth
from oauth2client.client import GoogleCredentials
# Authenticate and create the PyDrive client.
auth.authenticate_user()
gauth = GoogleAuth()
gauth.credentials = GoogleCredentials.get_application_default()
drive = GoogleDrive(gauth)
# https://drive.google.com/file/d/1S1w0Z7g3bI1PGLPR49PW5VBRo7c_KYgU/view?usp=sharing
link = 'https://drive.google.com/open?id=1S1w0Z7g3bI1PGLPR49PW5VBRo7c_KYgU'
fluff, id = link.split('=')
print (id) # Verify that you have everything after '='
downloaded = drive.CreateFile({'id':id})
downloaded.GetContentFile('Filename.csv')
import tensorflow as tf
ds = tf.data.experimental.make_csv_dataset('Filename.csv', batch_size=100, num_epochs=1)
iterator = ds.as_numpy_iterator()
print(next(iterator))
it works for me. Thanks for the interest to the topic (if somebody tried)
even simplier
# Load the Drive helper and mount
from google.colab import drive
drive.mount('/content/drive')
_types = [float(), float(), float(), float(), str()]
_lines = tf.data.TextLineDataset('/content/drive/My Drive/iris.csv')
ds=_lines.skip(1).map(lambda x: tf.io.decode_csv(x, record_defaults=_types) )
ds0= ds.take(2)
print(*ds0.as_numpy_iterator(), sep='\n') # print list with sep => by rows.
OR from df: (and batched for memory economical usage)
import tensorflow as tf
# Load the Drive helper and mount
from google.colab import drive
drive.flush_and_unmount()
drive.mount('/content/drive')
df= pd.read_csv('/content/drive/My Drive/iris.csv', dtype = 'float32', converters = {'variety' : str}, nrows=20, decimal='.')
ds = tf.data.Dataset.from_tensor_slices(dict(df)) # if mixed types
ds = ds.shuffle(20, reshuffle_each_iteration=False ) # for train.ds ONLY!
ds = ds.batch(batch_size=4)
ds = ds.prefetch(4)
# labels
label= ds.map(lambda x: x['variety'])
print(list(label.as_numpy_iterator()))
# features
#features = ds.map(lambda x: (x['sepal.length'], x['sepal.width']))
# Or with dynamic keys:
features = ds.map(lambda x: (list(map(x.get, list(np.setdiff1d(list(x.keys()),['variety']))))))
print(list(features.as_numpy_iterator()))
with any Transformations in map...
I have a sample spark dataframe that I create from pandas dataframe -
from pyspark.sql import SparkSession
import pyspark.sql.functions as F
from pyspark.sql.types import StringType
from pyspark.sql.types import *
import pandas as pd
spark = SparkSession \
.builder \
.appName("Python Spark SQL basic example") \
.config("spark.some.config.option", "some-value") \
.getOrCreate()
#create sample spark dataframe first and then create pandas dataframe from it
import pandas as pd
pdf = pd.DataFrame([[1,"hello world. lets shine and spread happiness"],[2,"not so sure"],[2,"cool i like it"],[2,"cool i like it"],[2,"cool i like it"]]
, columns = ['input1','input2'])
df = spark.createDataFrame(pdf) # this is spark df
now, I have the data types as
df.printSchema()
root
|-- input1: long (nullable = true)
|-- input2: string (nullable = true)
If i convert this spark dataframe back to pandas using -
pandas_df = df.toPandas()
and then if I try to print the data types, I get back object type for second column instead of string type.
pandas_df.dtypes
input1 int64
input2 object
dtype: object
How do I convert this string type in spark correctly to string type in pandas ?
To convert to string, you can use StringDtype:
pandas_df["input_2"] = pandas_df["input_2"].astype(pd.StringDtype())
I am trying to use Facebook prophet in spark in an Zeppelin environment and I have tried to follow the exact steps from https://github.com/facebook/prophet/issues/517, However, i get errors like below. I am simply not sure what am I to correct here or how to debug this.
My data contains a datetime features called ds, volume that I want to predict y and the segment and I am trying to build a model for each segment.
File"/hadoop14/yarn/nm/usercache/khasbab/appcache/application_1588090646020_2412/container_e168_1588090646020_2412_01_000001/py4j-0.10.7-src.zip/py4j/protocol.py", line 328, in get_return_value format(target_id, ".", name), value)
py4j.protocol.Py4JJavaError: An error occurred while calling o3737.showString.
%livycd.pyspark
from pyspark.sql.types import StructType,StructField,StringType,TimestampType,ArrayType,DoubleType
from pyspark.sql.functions import current_date
from pyspark.sql.functions import pandas_udf, PandasUDFType
from fbprophet import Prophet
from datetime import datetime
import pandas as pd
result_schema = StructType([
StructField('segment', StringType(), True),
StructField('ds', TimestampType(), True),
StructField('trend', ArrayType(DoubleType()), True),
StructField('trend_upper', ArrayType(DoubleType()), True),
StructField('trend_lower', ArrayType(DoubleType()), True),
StructField('yearly', ArrayType(DoubleType()), True),
StructField('yearly_upper', ArrayType(DoubleType()), True),
StructField('yearly_lower', ArrayType(DoubleType()), True),
StructField('yhat', ArrayType(DoubleType()), True),
StructField('yhat_upper', ArrayType(DoubleType()), True),
StructField('yhat_lower', ArrayType(DoubleType()), True),
StructField('multiplicative_terms', ArrayType(DoubleType()), True),
StructField('multiplicative_terms_upper', ArrayType(DoubleType()), True),
StructField('multiplicative_terms_lower', ArrayType(DoubleType()), True),
StructField('additive_terms', ArrayType(DoubleType()), True),
StructField('additive_terms_upper', ArrayType(DoubleType()), True),
StructField('additive_terms_lower', ArrayType(DoubleType()), True),
])
#pandas_udf(result_schema, PandasUDFType.GROUPED_MAP)
def forecast_loans(history_pd):
# instantiate the model, configure the parameters
model = Prophet(
interval_width=0.95,
growth='linear',
daily_seasonality=False,
weekly_seasonality=False,
yearly_seasonality=True,
seasonality_mode='multiplicative'
)
#history_pd['ds'] = pd.to_datetime(history_pd['ds'], errors = 'coerce', format = '%Y-%m-%d')
#.apply(lambda x: datetime.strptime(x,'%Y-%m-%d'))
# fit the model
model.fit(history_pd.loc[:,['ds','y']])
# configure predictions
future_pd = model.make_future_dataframe(
periods=20,
freq='W')
# make predictions
results_pd = model.predict(future_pd)
# return predictions
return pd.DataFrame({
'segment':history_pd['segment'].values[0],
'ds': [results_pd.loc[:,'ds'].values.tolist()],
'trend': [results_pd.loc[:,'ds'].values.tolist()],
'trend_upper': [results_pd.loc[:,'trend_upper'].values.tolist()],
'trend_lower': [results_pd.loc[:,'trend_lower'].values.tolist()],
'yearly': [results_pd.loc[:,'yearly'].values.tolist()],
'yearly_upper': [results_pd.loc[:,'yearly_upper'].values.tolist()],
'yearly_lower': [results_pd.loc[:,'yearly_lower'].values.tolist()],
'yhat': [results_pd.loc[:,'yhat'].values.tolist()],
'yhat_upper': [results_pd.loc[:,'yhat_upper'].values.tolist()],
'yhat_lower': [results_pd.loc[:,'yhat_lower'].values.tolist()],
'multiplicative_terms': [results_pd.loc[:,'multiplicative_terms'].values.tolist()],
'multiplicative_terms_upper': [results_pd.loc[:,'multiplicative_terms_upper'].values.tolist()],
'multiplicative_terms_lower': [results_pd.loc[:,'multiplicative_terms_lower'].values.tolist()],
'additive_terms': [results_pd.loc[:,'additive_terms'].values.tolist()],
'additive_terms_upper': [results_pd.loc[:,'additive_terms_upper'].values.tolist()],
'additive_terms_lower': [results_pd.loc[:,'additive_terms_lower'].values.tolist()]
})
#return pd.concat([pd.DataFrame(results_pd),pd.DataFrame(history_pd[['segment']].values[0])], axis = 1)
results =df3.groupBy('segment').apply(forecast_loans)
results.show()
I have tweaked my code to the following and downgraded to pyarrow 0.14 as suggested here Pandas scalar UDF failing, IllegalArgumentException and it all worked! I believe downgrading pyarrow to 0.14 was the key for spark 2.x versions as commented on stackoverflow.
The comment says the following "The issue is not with pyarrow's new release, it is spark which has to upgrade and become compatible with pyarrow.(i am afraid we have to wait for spark 3.0 to use the latest pyarrow)"
%livycd.pyspark
from pyspark.sql.types import StructType,StructField,StringType,TimestampType,ArrayType,DoubleType
from pyspark.sql.functions import current_date
from pyspark.sql.functions import pandas_udf, PandasUDFType
from fbprophet import Prophet
from datetime import datetime
import pandas as pd
result_schema = StructType([
StructField('segment', StringType(), True),
StructField('ds', TimestampType(), True),
StructField('trend', DoubleType(), True),
StructField('trend_upper', DoubleType(), True),
StructField('trend_lower', DoubleType(), True),
StructField('yearly', DoubleType(), True),
StructField('yearly_upper', DoubleType(), True),
StructField('yearly_lower', DoubleType(), True),
StructField('yhat', DoubleType(), True),
StructField('yhat_upper', DoubleType(), True),
StructField('yhat_lower', DoubleType(), True),
StructField('multiplicative_terms', DoubleType(), True),
StructField('multiplicative_terms_upper', DoubleType(), True),
StructField('multiplicative_terms_lower', DoubleType(), True),
StructField('additive_terms', DoubleType(), True),
StructField('additive_terms_upper', DoubleType(), True),
StructField('additive_terms_lower', DoubleType(), True),
])
#pandas_udf(result_schema, PandasUDFType.GROUPED_MAP)
def forecast_loans(df):
def prophet_model(df,test_start_date):
df['ds'] = pd.to_datetime(df['ds'])
# train
ts_train = (df
.query('ds < #test_start_date')
.sort_values('ds')
)
# test
ts_test = (df
.query('ds >= #test_start_date')
.sort_values('ds')
.drop('y', axis=1)
)
print(ts_test.columns)
# instantiate the model, configure the parameters
model = Prophet(
interval_width=0.95,
growth='linear',
daily_seasonality=False,
weekly_seasonality=False,
yearly_seasonality=True,
seasonality_mode='multiplicative'
)
# fit the model
model.fit(ts_train.loc[:,['ds','y']])
# configure predictions
future_pd = model.make_future_dataframe(
periods=len(ts_test),
freq='W')
# make predictions
results_pd = model.predict(future_pd)
results_pd = pd.concat([results_pd,df['segment']],axis = 1)
return pd.DataFrame(results_pd, columns = result_schema.fieldNames())
# return predictions
return prophet_model(df, test_start_date= '2019-03-31')
results =df3.groupBy('segment').apply(forecast_loans)
Assuming you are using Spark 2.3.x or 2.4.x and PyArrow >= 0.15.0, there is a known compatibility issue between PyArrow and Spark.
The simplest solution is to set the environment variable ARROW_PRE_0_15_IPC_FORMAT=1. The Spark documentation recommends setting it in conf/spark-env.sh, but you can set it in your Linux shell, and it is also possible to set it before creating your spark_session in your Python script or shell.
import os
os.ENVIRON["ARROW_PRE_0_15_IPC_FORMAT"] = "1"
spark_session = ...
Alternatively, you can downgrade PyArrow if that is an option for you, as noted in the other answer.