Szerkesztés

Megosztás a következőn keresztül:

Data Wrangling using PROSE Code Accelerator

  • Cikk

Applies to: SQL Server 2019 (15.x)

Important

The Microsoft SQL Server 2019 Big Data Clusters add-on will be retired. Support for SQL Server 2019 Big Data Clusters will end on February 28, 2025. All existing users of SQL Server 2019 with Software Assurance will be fully supported on the platform and the software will continue to be maintained through SQL Server cumulative updates until that time. For more information, see the announcement blog post and Big data options on the Microsoft SQL Server platform.

PROSE Code Accelerator generates readable Python code for your data wrangling tasks. You can mix the generated code with your hand-written code while working in a notebook within Azure Data Studio.

This article provides an overview of how you can use the Code Accelerator.

Note

Program Synthesis using Examples, aka PROSE, is a Microsoft technology that generates human-readable code using AI. It does so by analyzing a user's intent as well as data, generating several candidate programs, and picking the best program using ranking algorithms. To know more about the PROSE technology, visit the PROSE homepage.

The Code Accelerator comes pre-installed with Azure Data Studio. You can import it like any other Python package in the notebook. By convention, we import it as cx for short.

import prose.codeaccelerator as cx

In the current release, the Code Accelerator can intelligently generate Python code for the following tasks:

  • Reading data files to a Pandas or Pyspark dataframe.
  • Fixing data types in a dataframe.
  • Finding regular expressions representing patterns in a list of strings.

To get a general overview of Code Accelerator methods, see the documentation.

Read data from a file to a dataframe

Reading files to a dataframe involves looking at the content of the file and determining the correct parameters to pass to a data-loading library.

Depending on the complexity of the file, identifying the correct parameters might require several iterations.

PROSE Code Accelerator solves this problem by analyzing the structure of the data file and automatically generating code to load the file. Normally, the generated code parses the data correctly. In a few cases, you might need to tweak the code to meet your needs.

Consider the following example:

import prose.codeaccelerator as cx

# Call the ReadCsvBuilder builder to analyze the file content and generate code to load it
builder = cx.ReadCsvBuilder(r'C:/911.txt')

#Set target to pyspark if generating code to use pyspark library
#builder.Target = "pyspark"

#Get the code generated to fix the data types
builder.learn().code()

The previous code block prints the following python code to read a delimited file. Notice how PROSE automatically figures out the number of lines to skip, headers, quotechars, delimiters, etc.

import pandas as pd

def read_file(file):
   names = ["lat",
            "lng",
            "desc",
            "zip",
            "title"]

   df = pd.read_csv(file,
       skiprows = 11,
       header = None,
       names = names,
       quotechar = "\"",
       delimiter = "|",
       index_col = False,
       dtype = str,
       na_values = [],
       keep_default_na = False,
       skipinitialspace = True)
   return df

Code Accelerator can generate code to load delimited, JSON, and fixed-width files to a dataframe. For reading fixed-width files, the ReadFwfBuilder optionally takes a human-readable schema file that it can parse to get the column positions. To learn more, see the documentation.

Fix data types in a dataframe

It's common to have a pandas or pyspark dataframe with wrong data types. The incorrect data type happens because of a few non-conforming values in a column. As a result, Integers are read as Float or Strings, and Dates are read as Strings. The effort required to manually fix the data types is proportional to the number of columns.

You can use the DetectTypesBuilder in these situations. It analyzes the data and generates code to fix the data types. The code serves as a starting point. You can review, use, or modify it as needed.

import prose.codeaccelerator as cx

builder = cx.DetectTypesBuilder(df)

#Set target to pyspark if working with pyspark
#builder.Target = "pyspark"

#Get the code generated to fix the data types
builder.learn().code()

To learn more, see the documentation.

Identify patterns in Strings

Row Name BirthDate
0 Bertram du Plessis 1995
1 Naiara Moravcikova Unknown
2 Jihoo Spel 2014
3 Viachaslau Gordan Hilario 22-Apr-67
4 Maya de Villiers 19-Mar-60

Depending on the volume and diversity in data, writing regular expressions for different patterns in the column can be a very time consuming task. The FindPatternsBuilder is a powerful code acceleration tool that solves the above problem by generating regular expressions for a list of Strings.

import prose.codeaccelerator as cx

builder = cx.FindPatternsBuilder(df['BirthDate'])

#Set target to pyspark if working with pyspark
#builder.Target = "pyspark"

builder.learn().regexes

Here are the regular expressions generated by the FindPatternsBuilder for the above data.

^[0-9]{2}-[A-Z][a-z]+-[0-9]{2}$
^[0-9]{2}[\s][A-Z][a-z]+[\s][0-9]{4}$
^[0-9]{4}$
^Unknown$

Apart from generating Regular Expressions, FindPatternsBuilder can also generate code for clustering the values based on generated regexes. It can also assert that all the values in a column conform to the generated regular expressions. To learn more and see other useful scenarios, see the Microsoft PROSE Code Accelerator SDK.

Related content