Forum Discussion
Replacing multiple values in multiple columns
Hi
I need to replace multiple values in multiple columns based on a couple of translation tables.
The following solution provied values replacement on the entire table based on one translation table.
My question: is it possible, (based on the following solution or another) to control which column will be update i.e. -
Replace: Table [ Column1 ] values with 'Translation table1' values,
Table [ Column2 ] values with 'Ttranslation table2'....etc.
Here is MarcelBeug great solution for the entire table (without controling which column will be updated)
List.Accumulate(Table.ToRows(Translations),Table1,(t,r) => Table.ReplaceValue(t,r{0},r{1},Replacer.ReplaceValue,Table.ColumnNames(t)))http://community.powerbi.com/t5/Desktop/Power-Query-Batch-Replace-String/td-p/164402
Explanation:
- Translations is a table with translations (foreign - English), which is converted to a list of lists with replacements using Table.ToRows
- Table1 is your table. For List.Accumulate this table is the start value.
- The third argument for List.Accumulate is a function using a "state" and "current" parameter (i.c. t and r)
t is the table before each iteration and r is the foreign - English pair of the current iteration. - List.Accumulate will loop over the translation list and for each entry (or iteration), it will update your table (t) by replacing the old value (r{0}) ny the new value (r{1}), using function Replacer.ReplaceValue, and for all table columns (Table.ColumnNames(t)).
If each translation table is for 1 column, then I would suggest to use Table.TransformColumns instead of Table.ReplaceValues.
The query below has a double loop:
1. Outer loop over the columns in which values must be replaced (in the example: Name and Surname)
with the translation tables to use (in the example TranslationTable1 and -2),
2. Inner loop over each entry in the translation list.
r1 = Transformspecs, r1{0} = Column name to transform, r1{1} = translation list
t1/t2 = the table that is being transformed
r2 = translation list: r2{0} = old value, r2{1} = new value.
let ColumnsToTransform = {"Name", "Surname"}, TranslationLists = List.Buffer({Table.ToRows(TranslationTable1),Table.ToRows(TranslationTable2)}), TransformSpecs = List.Buffer(List.Zip({ColumnsToTransform,TranslationLists})), Source = Table1, Replaced = List.Accumulate( TransformSpecs, Source, (t1,r1) => List.Accumulate( r1{1}, t1, (t2,r2) => Table.TransformColumns( t2, {r1{0}, each Replacer.ReplaceValue(_,r2{0},r2{1})} ) ) ), Result = Value.ReplaceType(Replaced,Value.Type(Source)) in Result
2 Replies
If each translation table is for 1 column, then I would suggest to use Table.TransformColumns instead of Table.ReplaceValues.
The query below has a double loop:
1. Outer loop over the columns in which values must be replaced (in the example: Name and Surname)
with the translation tables to use (in the example TranslationTable1 and -2),
2. Inner loop over each entry in the translation list.
r1 = Transformspecs, r1{0} = Column name to transform, r1{1} = translation list
t1/t2 = the table that is being transformed
r2 = translation list: r2{0} = old value, r2{1} = new value.
let ColumnsToTransform = {"Name", "Surname"}, TranslationLists = List.Buffer({Table.ToRows(TranslationTable1),Table.ToRows(TranslationTable2)}), TransformSpecs = List.Buffer(List.Zip({ColumnsToTransform,TranslationLists})), Source = Table1, Replaced = List.Accumulate( TransformSpecs, Source, (t1,r1) => List.Accumulate( r1{1}, t1, (t2,r2) => Table.TransformColumns( t2, {r1{0}, each Replacer.ReplaceValue(_,r2{0},r2{1})} ) ) ), Result = Value.ReplaceType(Replaced,Value.Type(Source)) in Result
