In a complex workbook for twitter sentiment analysis I added a filter to reduce a list of tweets. Unexpectedly all my remaining tweets were rated positive. Instead of tweet analysis on a reduced list I got quite a different logic, negative word were counted positive. The graph in the details view showed quite a different execution plan, far less bars after the filtering, seems there is some optimization in the execution plan. For further analysis I switched on the save option for all worksheets. But that seems to repair the behaviour. So it seems like an optimization issue in the combination of this particular filter and storing only the result worksheets (the recommended practice).
Please sign in to leave a comment.