DataFrame in Expression Language

From QPR ProcessAnalyzer Wiki
Jump to navigation Jump to search

DataFrame represents a two dimensional array of data with one-to-many columns and zero-to-many rows, like a relational database table, an Excel sheet or a CSV data file. Each column in the DataFrame has a name, and there must not be more than one column with the same name. DataFrame is the generic data structure used to manage all kinds data in QPR ProcessAnalyzer expression engine that run in-memory. DataFrames as linked to other entities in QPR ProcessAnalyzer as follows:

  • Datatable contents is fetched into the memory as a DataFrame object
  • DataFrame can be stored (persisted) to a Datatable
  • ETL operations, such as joining, unions, filtering and grouping are based on the DataFrames
  • Data extracted from an external data source, e.g. using ODBC, is fetched to the in-memory calculation as a DataFrame.
  • When using a loading scripts, cases and events data is fed to the model using the DataFrame.
DataFrame properties Description
Columns (String*) DataFrame columns names as an array in the order the columns are in the DataFrame.
Rows (Object**) Returns the data content of the DataFrame as a two-dimensional array (matrix). The column names are not part of the data content.

Examples:

DatatableById(5).DataFrame.Rows[0][0]
Returns: the value in the first row and first column in a datatable with id 5.
<column name> (Object*)

Returns an array of values of given column in the datatable. If the column name contains spaces, the Column function needs to be used to refer to a column.

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).right
Returns: [zero, two, three]

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).right[2]
Returns: three
DataFrame functions Parameters Description
Append (DataFrame)
  • DataFrame which data to append

Creates a new DataFrame that has the contents of given DataFrame added to the end of this DataFrame. When the data is combined, the order of columns matters, not the names of the columns. The resulting DataFrame gets column names from this DataFrame.

If the number of columns is different between this DataFrame and the other DataFrame, an exception is thrown.

Examples:

Let("dataframe1", ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "text"]));
Let("dataframe2", ToDataFrame([[1, "one"], [4, "four"]], ["id", "text"]));
dataframe1.Append(dataframe2);

Returns:
id;text
0;zero
2;two
3;three
1;one
4;four
ColumnIndexes (Integer*)
  • Column names (String*)

Convert DataFrame column names into column indexes. The indexes are starting from zero. If a column is not found, an exception is given.

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).ColumnIndexes(["right", "id"])
Returns: [1, 0]
Column (Object*)
  • Column name

Returns an array of values of given column in the order rows are in the datatable.

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).Column("right")
Returns: [zero, two, three]
Columns (DataFrame)
  • Array of column names

Creates a new DataFrame having only the defined columns of the original DataFrame. Note that Columns function is different than Columns property (difference is that the function has parameters).

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).Columns(["right"]).ToCsv()
Returns:
right
zero
two
three

Head (DataFrame)
  • Number of rows

Creates a new DataFrame that only contains top number of rows of this DataFrame. If the DataFrame has less than n rows, all its rows are returned.

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).Head(2).ToCsv()
Results string:
id;right
0;zero
2;two
Join (DataFrame)
  1. Other DataFrame
  2. Columns to join
  3. join type

See Joining DataFrames.

GroupBy (DataFrame)
  1. Array of columns to group
  2. Array of grouping expressions

Creates a new DataFrame based on the current DataFrame. The resulting DataFrame has rows grouped by given columns and values aggregated using given functions. In the resulting DataFrame one row in the end result corresponds with one group.

Parameters:

  1. Columns: Columns to group identified by column names.
  2. Aggregation expressions: Array containing the column name as a key and the aggregation expression as a value.

Examples:

ToDataFrame([[0, "zero"], [0, "zero2"], [2, "two"], [2, "two"], [2, "two3"], [3, "three"]], ["id", "text"]).GroupBy(["id"],
[
  "ids": Def("", Sum(id)),
  "texts": Def("", StringJoin(",", text)),
  "constant": 123
]).ToCsv()
Returns:
ids;texts;constant
0;zero,zero2;123
6;two,two,two3;123
3;three;123

ToDataFrame([[0, "zero"], [0, "zero2"], [2, "two"], [2, "two"], [2, "two3"], [3, "three"]], ["id", "text"]).GroupBy(["id", "text"],
[
  "ids": Def("", Sum(id)),
  "texts": Def("", StringJoin(",", text)),
  "constant": 123
]).ToCsv()
Returns:
ids;texts;constant
0;zero;123
0;zero2;123
4;two,two;123
2;two3;123
3;three;123

Analysis("OperationLog")
.GroupBy(
  ["User Name"],
  [
    "User Name": Def("", Column("User Name")[0]),
    "Count": Def("Rows", CountTop(Rows)),
    "Avg. Duration": Def("", Average(Duration)),
    "Max. Duration": Def("", Max(Duration))
  ]
).ToCsv()

Returns (similar to this):
User Name;Count;Avg. Duration;Max. Duration
;207;0.617434782608696;20.556
Administrator;665;16.3750631578947;4225.497
qpr;128;2.158765625;20.346
Merge (DataFrame)

See Merging DataFrames.

OrderBy (DataFrame)
  • ordering expression

Creates a new DataFrame having rows ordered in an ascending order using the given expression evaluated on each row.

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "text"]).OrderBy(text).ToCsv()
Returns:
id;right
3;three
2;two
0;zero

Analysis("OperationLog").OrderBy(Duration).Head(1)
Results string:
id;right
0;zero
2;two
3;three
OrderByDescending (DataFrame)
  • ordering expression

Creates a new DataFrame having rows ordered in an descending order using the given expression evaluated on each row. See OrderBy above for examples.

Persist (DataTable)
  • DataTable name
  • Additional parameters

Writes a DataFrame into a DataTable in the QPR ProcessAnalyzer database. If a DataTable with that name does not exist in the project, a new DataTable is created. If a DataTable with that name already exists, the DataFrame will be stored into that DataTable. The function returns the written DataTable object.

The additional parameters support:

  • Append: Can be used to determine whether to append (true) or overwrite (false) the existing data. Default is false.
  • ProjectName: Name of the project under which the DataTable is to be created.
  • ProjectId: Id of the project under which the DataTable is to be created.

Examples:

Let("right", ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]));
right.Persist("RightDataTable", ["ProjectName": "TestData"])
Results: Id of the new data table named "RightDataTable" created into project named TestData (which is created if it doesn't already exist). If the table already existed, its contents will be overwritten by the new content.

Let("newData", ToDataFrame([[4, "four"]], ["id", "right"]));
newData.Persist("RightDataTable", ["ProjectName": "TestData", "Append": true])
Results: Id of the new data table named "RightDataTable" created into project named TestData (which is created if it doesn't already exist). If the table already existed, new content will be appended into the end of the table.
SetColumns (DataFrame)
  • new/modified columns as array

Creates a new DataFrame based on the current DataFrame, where new columns have been created and/or existing columns have been modified. New and modified columns are defined using an array, where the column name is as a key and as a value there is the expression to calculate the new or modified column. When specifying a column that already exists, the column values are modified. When specifying a new column name, that column is created as a new column to the resulting DataFrame.

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "text"]).SetColumns(["both": Def("", text + "=" + id)]).ToCsv()
Returns:
id;text;both
0;zero;zero=0
2;two;two=2
3;three;three=3

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "text"]).SetColumns(["text": Def("", text + "=" + id)]).ToCsv()
Returns:
id;text
0;zero=0
2;two=2
3;three=3

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "text"]).SetColumns(["both": Def("", text + "=" + id), "both+1": Def("", both + 1), "text": Def("", "Done: " + Column("both+1")), "constant": 1234]).ToCsv()
Returns:
id;text;both;both+1;constant
0;Done: zero=01;zero=0;zero=01;1234
2;Done: two=21;two=2;two=21;1234
3;Done: three=31;three=3;three=31;1234
Tail (DataFrame)
  • Number of rows

Creates a new DataFrame that has only the bottom number of rows of this DataFrame. If the DataFrame has less than n rows, all its rows are returned.

Example:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).Tail(2).ToCsv()
Results string:
id;right
2;two
3;three
ToCsv (String) (none)

Converts a DataFrame into a CSV data. The CSV data has the following formatting:

  • Column separator: semicolon
  • Decimal separator in numeric fields: period
  • Quotation character for text fields: double quotes (used when the textual value contains semicolon, double quotes, linebreak or tabulator)
  • Escape character: Double quotes in textual fields are escaped with two double quotes.
  • Date format for date fields: yyyy-MM-dd HH:mm:ss,fff
  • First line: contains column headers

Example:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).ToCsv()
Returns:
id;right
0;zero
2;two
3;three
Where (DataFrame)
  • condition expression

Creates a new DataFrame having only rows for which the given condition expression returns true. The condition expression can refer to the columns of the DataFrame (see the example below).

Examples:

Let("df", ToDataFrame([[0, "zero"], [2, "two", true], [3, "three"]], ["id", "string"]));

All the following expression return the same:
df.Where(id < 3);
df.Where(Column("id") < 3);
df.Where(_[0] < 3);

Returns:
id;string
0;zero
2;two
Zip (DataFrame)
  • DataFrame

Creates a new DataFrame that has the contents of given DataFrame appended as new columns into the end of this DataFrame. Returns a new DataFrame that has the colums from both the data frames so that the columns from the other DataFrame are appended to the end of the columns of this DataFrame. If the number of rows is different between this DataFrame and the other DataFrame, an exception is thrown. There must not be duplicate column names in the DataFrames - otherwise an exception is thrown.

Examples:

Let("df1", ToDataFrame([[0, "zero"], [1, "one"], [4, "four"]], ["id", "text"]));
Let("df2", ToDataFrame([[1, "one"], [2, "two"], [3, "three"]], ["id2", "text2"]));
df1.Zip(df2).ToCsv();
Returns:
id;text;id2;text2
0;zero;1;one
1;one;2;two
4;four;3;three

The following functions can be used to initialize DataFrame objects:

Function Parameters Description
ToDataFrame
  1. matrix
  2. column names array

Creates a DataFrame based on the given matrix (two dimensional array) and column names. Number of column names should be the same as the number of columns in the matrix.

Examples:

ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]).ToCsv()
Returns:
A string containing:
id;right
0;zero
2;two
3;three

Joining DataFrames

Performs a joining operation between two DataFrames.

Parameters:

  1. DataFrame: The other DataFrame to join.
  2. Columns to match: Columns which the joining is based on, can be defined as follows:
    • If joining using one column having the same name in both DataFrames, the column name is specified as as string.
    • If joining using several columns having the same names in both DataFrames, the column names are specified as a string array.
    • If joining using columns having different names between the DataFrames, columns are specified as an array of key-value pairs, where the key is the column name in the left side DataFrame, and value is the column name in the right side DataFrame.
  3. Join type: Join type can be inner (row is generated if both DataFrames have the key) or leftouter (at least one row is generated for each left side DataFrame row, even if there is no matching other row, in which case null is given as value for the other columns). The default is inner.

Examples:

Let("left", ToDataFrame([[0, "zero"], [1, "one"]], ["id", "left"]));
Let("right", ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]));
left.join(right, "id").ToCsv()
Returns:
id;left;right
0;zero;zero

Let("left", ToDataFrame([[0, "zero"], [1, "one"]], ["id", "left"]));
Let("right", ToDataFrame([[0, "zero"], [2, "two"], [3, "three"]], ["id", "right"]));
left.join(right, "id", "leftouter").ToCsv()
Returns:
id;left;right
0;zero;zero
1;one;

Let("left", ToDataFrame([[0, 0, "zerozeroleft"], [0, 1, "zeroleft"], [1, 2, "oneleft"]], ["idleft1", "idleft2", "left"]));
Let("right", ToDataFrame([[0, 0, "zerozeroright"], [0, 1, "zeroright"], [2, 3, "tworight"], [3, 4, "threeright"]], ["idright1", "idright2", "right"]));
left.join(right, ["idleft1": "idright1"], "inner");
Returns:
idleft1;idleft2;left;idright2;right
0;0;zerozeroleft;0;zerozeroright
0;0;zerozeroleft;1;zeroright
0;1;zeroleft;0;zerozeroright
0;1;zeroleft;1;zeroright

Let("left", ToDataFrame([[0, 0, "zerozeroleft"], [0, 1, "zeroleft"], [1, 2, "oneleft"]], ["idleft1", "idleft2", "left"]));
Let("right", ToDataFrame([[0, 0, "zerozeroright"], [0, 1, "zeroright"], [2, 3, "tworight"], [3, 4, "threeright"]], ["idright1", "idright2", "right"]));
left.join(right, ["idleft1": "idright1", "idleft2": "idright2"], "inner");
Returns:
idleft1;idleft2;left;right
0;0;zerozeroleft;zerozeroright
0;1;zeroleft;zeroright

Merging DataFrames

Creates a new DataFrame that has the contents of a DataFrame (target) merged with another DataFrame (source). The merging works with the same principle as in the SQL language. Note that the merging does not create new columns but the result DataFrame has the same columns as the target DataFrame.

Parameters:

  1. Source DataFrame: DataFrame to be merged with the target DataFrame.
  2. Columns to MATCH: Columns used to match the source and target DataFrames. This parameter has similar syntax as the 2. parameter of the join function.
  3. Columns to UPDATE when matching rows: Columns to copy from the source to target for the matching rows. This parameter is defined like the join function's matchColumns parameter (2. parameter).
    • If the value is _remove, all matching rows are be removed from the resulting table.
    • If the value is null, all rows are copied from the source to target (the ones having matching names). The default value is null.
    • If the value is [], no columns are updated.
  4. Columns to CREATE for non-match rowsing: Columns to create to target, if a matching row is not found. This parameter is defined like the previous update parameter.
  5. Keep or DELETE if not matched by source: Boolean value that defines whether those rows in the target DataFrame where no matching row in the source DataFrame is found, are included in the result (true) or deleted (false). The default value is true.

Examples:

Update case attribute values for some cases and create non-existing:

Let("CaseData", ToDataFrame([
	["0", "New York", 6],
	["1", "Dallas", 3],
	["2", "Dallas", 8],
	["3", "Chicago", 4],
	["4", "New York", 2]
],["Case id", "Region", "Cost"]));
Let("UpdatedData", ToDataFrame([
	["0", 7],
	["1", 5],
	["4", 2],
	["5", 2]
],["Case id", "Cost"]));
CaseData.Merge(UpdatedData, "Case id").tocsv();

Returns:
Case id;Region;Cost
0;New York;7
1;Dallas;5
2;Dallas;8
3;Chicago;4
4;New York;2
5;;2

Update case attribute values for some cases and not create non-existing:

Let("CaseData", ToDataFrame([
	["0", "New York", 6],
	["1", "Dallas", 3],
	["2", "Dallas", 8],
	["3", "Chicago", 4],
	["4", "New York", 2]
],["Case id", "Region", "Cost"]));
Let("UpdatedData", ToDataFrame([
	["0", 7],
	["1", 5],
	["4", 2],
	["5", 2]
],["Case id", "Cost"]));
CaseData.Merge(UpdatedData, "Case id", null, _remove).tocsv();

Returns:
Case id;Region;Cost
0;New York;7
1;Dallas;5
2;Dallas;8
3;Chicago;4
4;New York;2

Update case attribute values for some cases (create non-existing) where columns to match have different names in the source and target DataFrames ("Case id" in target and "Case" in source are used to match, and "Cost" in target is updated from "Variable Cost" in source):

Let("CaseData", ToDataFrame([
	["0", "New York", 6],
	["1", "Dallas", 3],
	["2", "Dallas", 8],
	["3", "Chicago", 4],
	["4", "New York", 2]
],["Case id", "Region", "Cost"]));
Let("UpdatedData", ToDataFrame([
	["0", 7, 4],
	["1", 5, 2],
	["4", 2, 0],
	["5", 2, 1]
],["Case", "Cost", "Variable Cost"]));
CaseData.Merge(UpdatedData, ["Case id": "Case"], ["Cost": "Variable Cost"]).tocsv();

Returns:
Case id;Region;Cost
0;New York;4
1;Dallas;2
2;Dallas;8
3;Chicago;4
4;New York;0
5;;1

Delete matching cases (don't create non-matching):

Let("CaseData", ToDataFrame([
	["0", "New York", 6],
	["1", "Dallas", 3],
	["2", "Dallas", 8],
	["3", "Chicago", 4],
	["4", "New York", 2]
],["Case id", "Region", "Cost"]));
Let("UpdatedData", ToDataFrame([
	["0"],
	["1"],
	["4"],
	["5"]
],["Case id"]));
CaseData.Merge(UpdatedData, "Case id", _remove, _remove).tocsv();

Returns:
Case id;Region;Cost
2;Dallas;8
3;Chicago;4

Update matching cases, create non-matching by source as new, and delete non-matching by target:

Let("CaseData", ToDataFrame([
	["0", "New York", 6],
	["1", "Dallas", 3],
	["2", "Dallas", 8],
	["3", "Chicago", 4],
	["4", "New York", 2]
],["Case id", "Region", "Cost"]));
Let("UpdatedData", ToDataFrame([
	["0", 7],
	["1", 5],
	["4", 2],
	["5", 2]
],["Case id", "Cost"]));
CaseData.Merge(UpdatedData, "Case id", null, null, false).tocsv();

Returns:
Case id;Region;Cost
0;New York;7
1;Dallas;5
4;New York;2
5;;2
Let("target", ToDataFrame([[0, "zero", "target"], [1, "", "target"]], ["id", "text", "frame"]));
Let("source", ToDataFrame([[1, "one", "source"], [2, "two", "source"], [3, "three", "source"]], ["id", "text", "frame"]));

target.Merge(source, "id").ToCsv()
Returns (one key, default parameters, identical dataframe columns):
id;text;frame
0;zero;target
1;one;source
2;two;source
3;three;source

target.Merge(source, "id", ["text"]).ToCsv()
Returns (one key, default parameters, identical dataframe columns, copy only text column from source):
id;text;frame
0;zero;target
1;one;target
2;two;
3;three;

target.Merge(source, "id", ["text"], _remove).ToCsv()
Returns (one key, default parameters, identical dataframe columns, copy only text column from source, remove rows found only in source):
id;text;frame
0;zero;target
1;one;target

target.Merge(source, "id", ["text"], _remove, false).ToCsv()
Returns (one key, identical dataframe columns, copy only text column from source, remove rows found only in source or only in target):
id;text;frame
1;one;target

target.Merge(source, "id", _remove, _remove, false).ToCsv()
Returns (one key, identical dataframe columns, remove all rows):
id;text;frame


Let("target", ToDataFrame([[0, 0, "zerozeroleft", "target"], [0, 1, "zeroleft", "target"], [1, 2, "left", "target"], [4, 5, "fourleft", "target"]], ["idleft1", "idleft2", "textleft", "frame"]));
Let("source", ToDataFrame([[0, 0, "zerozeroright", "source"], [0, 1, "zeroright", "target"], [1, 2, "oneright", "source"], [2, 3, "tworight", "source"], [3, 4, "threeright", "source"]], ["idright1", "idright2", "textright", "frame"]));

target.Merge(source, ["idleft1": "idright1"]).ToCsv()
Returns (one key, default parameters, different dataframe columns, copy all matching columns):
idleft1;idleft2;textleft;frame
0;0;zerozeroleft;source
0;0;zerozeroleft;target
0;1;zeroleft;source
0;1;zeroleft;target
1;2;left;source
4;5;fourleft;target
2;;;source
3;;;source

target.Merge(source, ["idleft1": "idright1"], []).ToCsv()
Returns (one key, default parameters, different dataframe columns, copy only key column):
idleft1;idleft2;textleft;frame
0;0;zerozeroleft;target
0;0;zerozeroleft;target
0;1;zeroleft;target
0;1;zeroleft;target
1;2;left;target
4;5;fourleft;target
2;;;
3;;;

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"]).ToCsv()
Returns (two keys, default parameters, different dataframe columns, copy all matching columns):
idleft1;idleft2;textleft;frame
0;0;zerozeroleft;source
0;1;zeroleft;target
1;2;left;source
4;5;fourleft;target
2;3;;source
3;4;;source

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"], ["textleft": "textright"]).ToCsv()
Returns (two keys, default parameters, different dataframe columns, copy only textright -column):
idleft1;idleft2;textleft;frame
0;0;zerozeroright;target
0;1;zeroright;target
1;2;oneright;target
4;5;fourleft;target
2;3;tworight;
3;4;threeright;

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"], ["textleft": "textright", "frame"]).ToCsv()
Returns (two keys, default parameters, different dataframe columns, copy textright and frame -columns):
idleft1;idleft2;textleft;frame
0;0;zerozeroright;source
0;1;zeroright;target
1;2;oneright;source
4;5;fourleft;target
2;3;tworight;source
3;4;threeright;source

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"], _remove).ToCsv()
Returns (two keys, default parameters, different dataframe columns, remove all matching rows, copy only matching columns):
idleft1;idleft2;textleft;frame
4;5;fourleft;target
2;3;;
3;4;;

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"], ["textleft": "textright"], ["idleft1": "idright1", "idleft2": "idright2", "textleft": "textright"]).ToCsv()
Returns (two keys, default parameters, different dataframe columns, copy only textright-column for matching columns, copy id columns and textright-column for rows not found in target):
idleft1;idleft2;textleft;frame
0;0;zerozeroright;target
0;1;zeroright;target
1;2;oneright;target
4;5;fourleft;target
2;3;tworight;
3;4;threeright;

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"], ["textleft": "textright"], ["idleft1": "idright1", "idleft2": "idright2", "textleft": "textright", "frame"]).ToCsv()
Returns (two keys, default parameters, different dataframe columns, copy only textright-column for matching columns, copy id, frame and textright-column for rows not found in target):
idleft1;idleft2;textleft;frame
0;0;zerozeroright;target
0;1;zeroright;target
1;2;oneright;target
4;5;fourleft;target
2;3;tworight;source
3;4;threeright;source

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"], null, ["idleft1": "idright1", "idleft2": "idright2", "textleft": "textright"]).ToCsv()
Returns (two keys, default parameters, different dataframe columns, don't copy any columns from source for matching columns, copy id columns and textright-column for rows not found in target):
idleft1;idleft2;textleft;frame
0;0;zerozeroleft;source
0;1;zeroleft;target
1;2;left;source
4;5;fourleft;target
2;3;tworight;
3;4;threeright;

target.Merge(source, ["idleft1": "idright1", "idleft2": "idright2"], ["textleft": "textright"], ["idleft1": "idright1", "idleft2": "idright2", "textleft": "textright", "frame"], false).ToCsv()
Returns (two keys, default parameters, different dataframe columns, copy only textright-column for matching columns, copy id, frame and textright-column for rows not found in target, remove all rows not found in source):
idleft1;idleft2;textleft;frame
0;0;zerozeroright;target
0;1;zeroright;target
1;2;oneright;target
2;3;tworight;source
3;4;threeright;source