SqlDataFrame in Expression Language

From QPR ProcessAnalyzer Wiki
Revision as of 21:49, 5 June 2023 by Ollvihe (talk | contribs)
Jump to navigation Jump to search

SqlDataFrame represents tabular data similar to an SQL query result. Data managed using SqlDataFrames is processed in the datasource where the data is located, i.e., in Snowflake or SQL Server (not in QPR ProcessAnalyzer Server memory). For each SqlDataFrame, there is an SQL query generated that is run in the datasource where the referenced datatables are located.

SqlDataFrame operations itself don't cause the SQL to execute in the datasource, but it will happen when the Collect function is called for an SqlDataFrame, which generates and executes the SQL query representing the SqlDataFrame and loads the data into QPR ProcessAnalyzer memory as a DataFrame (where it can be presented in a dashboard).

Each SqlDataFrame contain information about the datasource where the SQL query will be executed. When writing queries with several SqlDataFrames, the SqlDataFrames need to be located in the same datasource, to be able to execute the queries. If needed, data can be moved between datasources by using the Import or Persist functions. Alternatively, processing can be done in-memory by calling Collect for an SqlDataFrame and continuing calculation as in-memory DataFrame.

There is a similar API for the SqlDataFrames as there is for the DataFrames. Note that merging is not possible between SqlDataFrames like it's for DataFrames, but SqlDataFrame can be merged into a datatable.

SqlDataFrame properties

Property Description
ColumnTypes (Dictionary)

Returns information about the columns of the SqlDataFrame as an array of dictionaries with keys Name and DataType. Columns are returned in the same order as the columns exist in the data table. Column types are calculated based on the type of the column in the relational database management system (i.e. type of the column in an SQL table).

Examples:

table.SqlDataFrame.ColumnTypes
Returns: [ #{ "Name": "name1", "DataType": "Integer" },  #{ "Name": "name2", "DataType": "String" } ]

table.SqlDataFrame.ColumnTypes.Name
Returns: ["name1", "name2"]

table.SqlDataFrame.ColumnTypes.DataType
Returns: ["Integer", "String"]
NColumns (Integer) Returns number of columns in the dataset represented by the SqlDataFrame.
NRows (Integer) Returns number of rows in the dataset represented by the SqlDataFrame.

SqlDataFrame functions

Function Parameters Description
Aggregate (SqlDataFrame)
  1. Aggregated columns (string array or key-value pairs)
  2. Aggregation methods (string array)
Same functionality as in the DataFrame.
Append DataFrame to append Same functionality as in the DataFrame.
ApplyFilter (SqlDataFrame)
  1. Filter object
  2. SqlDataFrame for cases

Filters the SqlDataFrame (assumed to contain events data) using given filter and returns SqlDataFrame containing the remaining event data after filtering is performed. Requires that the source SqlDataFrame has CaseId, EventType and TimeStamp mappings defined.

Parameters:

  1. filter: Filter object.
  2. cases (optional): SqlDataFrame containing corresponding case data. Case data should have the CaseId mapping defined.

Example: Returns SqlDataFrame containing only events for case id "12345".

let model = ModelById(123);
model
  .EventsDataTable
  .SqlDataFrame
  .ApplyFilter(
    #{
      "Items": [ #{
        "Type": "IncludeCases",
        "Items": [ #{
            "Type": "Case",
            "Values": [ "12345" ]
        }]
      }]
    },
  model.CasesDataTable.SqlDataFrame
)
Collect (SqlDataFrame) (none)

Executes the SQL query for the SqlDataFrame in the datasource and returns results as an in-memory DataFrame. Then processing of the data can be continued as the in-memory DataFrame.

Examples:

DataTableById(123).SqlDataFrame.Head(100).Collect().ToCsv()
Returns the top 100 rows from datatable id 123.
ExcludeValues (SqlDataFrame)
  1. Column name (string)
  2. Value (single item) or values (array) to exclude

Same functionality as in the DataFrame.

GroupBy (GroupedDataFrame)

Grouped columns (string array)

Same functionality as in the DataFrame.

Head (SqlDataFrame) Number of top rows

Same functionality as in the DataFrame.

IncludeOnlyValues (SqlDataFrame)
  1. Column name (string)
  2. Value (single item) or values (array) to include

Same functionality as in the DataFrame.

Join (SqlDataFrame)
  1. DataFrame
  2. Columns to match (String or key-value pairs)
  3. Join type (String)
Same functionality as in the DataFrame.
OrderByColumns (SqlDataFrame)
  1. Ordered columns (String array)
  2. Sorting order (boolean array)

Same functionality as in the DataFrame.

Persist (SqlDataFrame)
  1. DataTable name
  2. Additional parameters

Same functionality as in the DataFrame. Additionally, if the SQL query for the SqlDataFrame is run in the same system as the target datatable, all data processing and storage is done within the system to achieve efficient operation.

Pivot (SqlDataFrame)
  1. Pivot column name (String)
  2. Value column name (String)
  3. Pivot values (String array)
  4. Aggregate function (String)

Performs the pivot operation for the dataframe. More information about pivot: https://www.techonthenet.com/sql_server/pivot.php.

Parameters:

  1. Pivot column name: Column in the source dataframe that will be aggregated.
  2. Value column name: Column in the source dataframe that determines to which generated column the row belongs to.
  3. Pivot values: Array of column names that the pivot operate will generate.
  4. Aggregate function: Aggregate function for combining the values from pivot column (the first parameter).

Example:

let data = ModelById(1)
.EventsDatatable
.SqlDataFrame;

let result1 = data
.Select(["User", "Event type"])
.Pivot("User", "User", ["John", "Greg", "James", "Sharon"], "Count").Collect();
// Returns user counts (on columns) for each event type (on rows)

let result2 = data
.Select(["User", "Event type", "Cost"])
.Pivot("Cost", "User", ["John", "Greg", "James", "Sharon"], "Sum").Collect();
// Returns costs for each user (on columns) and event type (on rows)
RemoveColumns (SqlDataFrame) Column names (string array) Same functionality as in the DataFrame.
Select (SqlDataFrame) Column names (string array, or key-value pairs) Same functionality as in the DataFrame.
SelectDistinct (SqlDataFrame)

Column names (string array, or key-value pairs)

Same functionality as in the DataFrame.

Skip (SqlDataFrame) Number of rows to skip Same functionality as in the DataFrame.
TakeSample (SqlDataFrame) Number of rows (Integer) Same functionality as in the DataFrame.
Unpivot (SqlDataFrame)
  1. Value column name (String)
  2. Name column name (String)
  3. Columns to unpivot (String array)

Performs the unpivot operation for the dataframe, i.e., rotates columns into rows. More information about unpivot: https://docs.snowflake.com/en/sql-reference/constructs/unpivot.html.

Parameters:

  1. Value column name: Name of the generated column containing the unpivotted values.
  2. Name column name: Name of the generated column containing original column names of the unpivotted values.
  3. Columns to unpivot: Names of the columns in the source dataframe to be unpivotted.

Note that the Unpivot function is only supported for SqlDataFrames (not in-memory DataFrames).

Example:

let result = df.Unpivot(
  "Value",
  "Name",
  ["Column 1", "Column 2", "Column 3"]
).Collect();

This example reads case attributes from a model and performs unpivot for them. Only string type of columns are unpivotted and also the case id columns is ignored.

let caseAttributes = ModelById(123).CasesDatatable;
caseAttributes.SqlDataFrame.Unpivot(
  "Value",
  "Case attribute",
  caseAttributes.Columns.Where(Datatype == "String" && Name != "CaseId").Name
).Collect().toCsv()
WithClusterColumn (SqlDataFrame)
  1. New column name (String)
  2. Parameters dictionary

Creates new SqlDataFrame with a new column whose value represents a cluster label generated using a clustering algorithm. Cluster labels are integer numbers starting from 0.

Parameters:

  1. columnId: Name/id of the new column.
  2. configuration: Configurations for the clustering algorithm.
    • NClusters: Number of clusters to split the data into.
    • Columns: Names of the columns to use in the clustering.

Uses K-Means algorithm implemented with scikit-learn KMeans python package where random_state is set to be 0. Different data types are preprocessed before performing the K-Means algorithm as follows:

  • Integers and floats are converted into numeric values between 0 and 1 using sklearn.preprocessin.MinMaxScaler (default settings).
  • Timestamps are converted into integer values, after it will get the same treatment as the numeric types described above.
  • All the other data types are converted into multiple one-hot-encoded columns using sklearn.preprocessing.OneHotEncoder.

Example: Returns a CSV representation of a dataframe whose rows have been clustered into three clusters by "Region" and "Accoung Manager" columns.

DataTableById(1)
  .SqlDataFrame
  .WithClusterColumn("Cluster", #{"NClusters": 3, "Columns": ["Region", "Account Manager"]})
  .Collect().ToCsv()
WithDenseRankNumberColumn (SqlDataFrame)
  1. New column name (String)
  2. Order by columns (String array)
  3. Partition by columns (String array)
  4. Ascending/descending order (Boolean array)

Same functionality as in the DataFrame.

WithColumn (SqlDataFrame)
  1. New column name (String)
  2. New column expression

Same functionality as in the DataFrame, except instead of in-memory expressions, SqlDataFrame use SQL Expressions.

WithRankColumn (SqlDataFrame)
  1. New column name (String)
  2. Order by columns (String array)
  3. Partition by columns (String array)
  4. Ascending/descending order (Boolean array)

Same functionality as in the DataFrame.

WithRowNumberColumn (SqlDataFrame)
  1. New column name (String)
  2. Order by columns (String array)
  3. Partition by columns (String array)
  4. Ascending/descending order (Boolean array)

Same functionality as in the DataFrame.

Where (SqlDataFrame) Condition expression Same functionality as in the DataFrame, except instead of in-memory expressions, SqlDataFrame use SQL Expressions.