I have a table with thousands of columns with numbers that I need to transpose to select and filter the values.Unfortunantely transposing a table leads to the loss of formats. How can I convert the formats of the colums to integer? This is needed so that I can write the table as a XLS-file
if you transpose a table only containing integers, they will result in integer columns again. However if you have other columns, like identifiers in there, those are in the same transposed column and they are no longer "pure" integer columns.
So, two kind of approaches:
1. Filter all non integer columns and than do the transposing.
2. Try the "Column Auto Type Cast" node. This converts the columns to the best column type.
Let me know if this works for you!
Option 1 works, but I lose important textfields. But with a separate table I can deal with it.
Option 2 makes strings of the numbers
Additonal question: how can I add column names?
but than transposing, is not what you should do? What is your final analysis goal?
The Former RowIds are afterwards your new column names. You can change RowIds with the RowId node, or you afterwards use a column rename node.
Best regards, Iris
The goal is to compare and examine long wordvectors, a product of a textmining workflow. I have tables with thousands of columns and a limited number of rows. By transposing this I create the opposite and then I can use sort descending to compare the most important terms in the wordvectors. But I need a characteristic (filename and category/class) to see which columns I compare. If I add these characteristics (strings) to the file the format is not reckognized. The order then by sorting descending is f.i 8.0-2.0-1.0-12.0. But I of course want the 12.0 before the 8.0. If I transpose without strings the format of the columns is D (standard double) and the sorting is correct but unfortunately I can recognize the columns only by order as a characteristic is missing. Hence my question to add a columnname afterwards. This is possible by using column rename but then I have to do this manual and I have to do this again when the wordvectors change. A column rename based on a referencetable should help but that isn't available I suppose.
I also tried depivotize but this doesn't do what I want. Hope this makes clear what I need.
Hi Iris, on your request I added more information, expecting more advice, but unfortunately I don't get a reply?
sorry about that. But in this case you are right, pivoting is the thing to do and you need to use two tables. One with the vectors, which you can transpose afterwards and one with the remaining informations.
The node for column renaming based on a reference table is the insert column header node.