Replace Groups of Values
Whether data is missing, mismatched, or simply wrong, you can use a variety of methods in the Cloud Portal to replace values in one or more columns with literal values or pattern-based replacements.
Replacement methods
In the Transformer page, you can use the following methods to replace values:
Method | Description |
|---|---|
By selection | Select a value in the data grid to prompt a series of suggestions on what to do with the data. Typically, replacement options are near the top of the suggestions. ヒント You can replace specific values in a column with a preferred value. For more information, see Replace Cell Values. |
By column menu | From the column menu, select Replace and a sub-menu item to begin configuring a replacement transformation. |
By Transformer toolbar | At the top of the data grid, click the Replace icon in the Transformer toolbar to begin configuring replacements. |
By Search panel | In the Search panel, enter |
Replace by selection
When you select data in the data grid, the replacement suggestions are pre-specified for you, including a number of variants available in the suggestion card.
Notes:
Suggestions are typically conservative in the scope of their changes. Case-sensitive searches and matching of the first occurrence only are the default settings.
Order of listing of suggestions in a suggestion card:
Pattern-based replacements are listed first. These replacements use Wrangle , instead of regular expressions. Regular expressions can be more difficult to control.
Literal value replacements are listed below the pattern-based ones.
For more information, seeOverview of Predictive Transformation.
Mask data
For privacy reasons or for sensitivity reasons, you may wish to mask sensitive data in one or more columns with fixed strings.
Delete whole column(s)
If you need to remove the data in an entire column, the easiest method is to delete a column. Select one or more columns and then select Delete from the column drop-down. See Remove Data.
Masking all values
You can use a transformation like the following to replace all values in a column with a simple string. In this case, the value #REDACTED# has been inserted in place of all values in the column.
注記
This replacement changes the data type of the column to String. If you must retain the original data type, the replacement value should be valid for the data type.
Transformation Name | |
|---|---|
Parameter: Columns | transactionValue |
Parameter: Formula | '#REDACTED#' |
Partial masking of values
Suppose you wish to partially mask data in a column. In the following example, data for the AcctNum column is masked, except for the last four characters (digits):
Transformation Name | |
|---|---|
Parameter: Columns | AcctNum |
Parameter: Formula | value: merge(['XXXX',right(AcctNum, 4)], '') |
Mask multiple columns based on data type
You can use the following type of transformation to hide data based on data type. In this example, the values in all columns with Social Security Number (SSN) are replaced with a masking value: XXX-XX-XXXX:
警告
This method performs a simple text replacement of the data in the columns(s). After this transformation has been applied to the data, the source data is no longer available, unless you step back to a step before this one. For these kinds of operations, you may find it more secure to apply these kinds of masking operations to the source data in a single recipe and then make that output available to other users to use as an imported dataset.
Transformation Name | |
|---|---|
Parameter: Columns | All |
Parameter: Formula | if(isvalid($col, ['SSN']), 'XXX-XX-XXXX', $col) |
Replace with values from another column
Replace whole column
You can do simple replacements of data from one column into another with transformations like the following. In this example, the values of colB are replaced with the values of colA with 0.15 added to them:
Transformation Name | |
|---|---|
Parameter: Columns | colB |
Parameter: Formula | colA + 0.15 |
Replace partial values from another column
You can use the MERGE function to blend full or partial sets of columns into a new column. In the following example, the newBrandId value is concatenated with the product code in the ProdId column to create a new product identifier:
Transformation Name | |
|---|---|
Parameter: Columns | ProdId |
Parameter: Formula | merge([newBrandId, right(prodId, 5)], '-') |
Replace between positions
You can perform replacements based on character positions that you specify as part of the transformation.
The beginning character value is specified as a number from 0, which starts on the left.
The ending character value must be equal to or greater than the beginning character value.
In the following example, the Whse_Name column values are prepended with the value old-.
Transformation Name | |
|---|---|
Parameter: Column | Whse_name |
Parameter: Start position | 0 |
Parameter: End position | 0 |
Parameter: Replace with | old- |
Search and replace text or pattern
You can search and replace content in your dataset based on literals or patterns. In the following example, the value ##CLT_NAME## is replaced with Our Customer, Inc. across all columns in the dataset:
Transformation Name | |
|---|---|
Parameter: Column | All |
Parameter: Find | '##CLT_NAME##' |
Parameter: Replace with | 'Our Customer, Inc.' |
Parameter: Match all occurrences | true |
Replace missing values
Replace missing with zeroes
For numeric data, you may choose to replace values that are missing in a column with zeros. The following transformation sets missing values in the Qty and DiscountPct columns of Decimal data type to 0:
Transformation Name | |
|---|---|
Parameter: Columns | Qty,DiscountPct |
Parameter: Formula | if(ismissing([$col]), '0', $col) |
Replace missing with average values
One of the problems with the above method is that any statistical computations applied to the column are now affected by the zeroing of the missing values. For example, the computation for the AVERAGE function does not factor in missing values into the count of rows, which result in skewing of values for your purposes.
The following example creates a new column from the DiscountPct column in which empty values are inserted as the average of the values in the source column:
Transformation Name | |
|---|---|
Parameter: Formula type | Single row formula |
Parameter: Formula | if(ismissing([DiscountPct]), average(DiscountPct), DiscountPct) |
Parameter: New column name | DiscountPct-0toAVG |
In this manner, the new column can be used for some statistical modeling, while preserving the original values in the original column.
Replace mismatched values
You can perform replacements based on the values in a column that are mismatched against a specified type.
In the following example, Datetime values that do not match the yyyy*mm*dd, where the asterisk (*) is a wildcard value.
Transformation Name | |
|---|---|
Parameter: Columns | Multiple |
Parameter: Column 1 | myDate |
Parameter: Data type to evaluate | Date/Time |
Parameter: Date/Time type | yyyy*mm*dd |
Parameter: Replace with | Custom value |
Parameter: New value | '##BAD_DATE##' |
注記
In the above example, the Date/Time type parameter applies only to replacements that are mismatched against the Date/Time data type. This parameter is used to specify the Datetime format against which the source values are validated. The parameter does not appear in Replace mismatched values transformations for other data types.