Data Mash Ups
Data mashups are simple when using the 1010data spreadsheet-like interface or any of the other available interfaces. The user simply joins two tables based on the values in one or more joining columns, even if there is no predefined relational model between the tables. The columns can already exist in the two tables or can be computed dynamically, allowing for complex fuzzy matching based on algorithms.
We tried to match data from DIY sales tables to commercial sales tables by store/item/date (350MM rows for 1 yr.). The only way we were able to meet the request (after weeks of SAS bombing out) was to write a SAS macro to loop through by store/date. This ran each fiscal year in a separate query and took about a week to complete (just for the code to run!). Once I had the query written in 1010data format, I was able to get results for one fiscal year in less than 30 minutes. |
1010data also supports innovative joins like "as-of joins", that allow inexact match ups between two or more time-series data sets.