I want to perform a join between two datasets, which by iteself isn't so hard. Unfortunately, I need to build one of the datasets on the fly and preferably in memory. I need to read a series of key values from an external binary file, join them with a table, and return a dataset where the binary file key values = table key values. Currently I'm reading the binary file one record at a time and performing a join with the table each time. So, for 865 records I'm executing the query 865 times. I'd like to put the keys into a table in memory (maybe a clientdata set?) and then just execute a single query. Does that sound possible?