Lloyd Tabb on Symmetric Aggregates
Download as PPTX, PDF•1 like•747 views
The document discusses the challenges of computing SQL aggregates, such as errors in sum() and avg() due to joins and complex queries. It introduces LookML, which simplifies model building by logically combining relation and aggregation, resulting in generalized aggregates that function across various queries. This approach eliminates incorrect results from joins and enables new analytical patterns and ratio calculations within the same explore.
Lloyd Tabb on Symmetric Aggregates
- 1. Symmetric Aggregates Look & Tell: NY March 24, 2015
- 2. SQL Aggregates are Error Prone ● When computing SUM() and AVG() in SQL its easy to make mistakes ● Adding a join can make computations incorrect ● Difficult to track down problems ● Leads to lots of separate queries.
- 3. SQL Queries Are Relate then Aggregate ● First: Relate and Filter (JOIN and WHERE clauses) ● Second: Aggregate (SUM and GROUP BY)
- 4. LookML Combines Relate and Aggregate ● Logically, LookML combines them so aggregates are much more generalized. ● SUM and, AVERAGE work in any query, regardless of relationship
- 5. So What? ● Radically simplifies model building (declare measures in their views and they just work). ● Joins don’t cause incorrect results. ● Fantastic new analytical patterns are now possible. ● Reduce the number of explores (base_views)
- 6. Patterns ● New Analytic Ratios ● Revenue and expense in the same explore ● Attribute based on analysis ● Paired behavior (i.e., bought together) ● Sloppy joined dates for trends
- 7. Upgrading Your Model ● Declare your relationship types in your joins. ● Turn it on!
- 8. Questions?
Editor's Notes
- #2: Intro yourself
- #3: Relating is the power of SQL. Very complex relationships can be described. Its all about matrix transformation. Most people don’t understand the power of this complex relating. Its about combining multiple tables (spreadsheets) into a single spreadsheet. Aggregating is the process of taking that spreadsheet and reducing it, folding along dimensions and computing measures. Window functions are a post processing step to produce things like rank and cummulative sums, etc.
- #4: Relating is the power of SQL. Very complex relationships can be described. Its all about matrix transformation. Most people don’t understand the power of this complex relating. Its about combining multiple tables (spreadsheets) into a single spreadsheet. Aggregating is the process of taking that spreadsheet and reducing it, folding along dimensions and computing measures. Window functions are a post processing step to produce things like rank and cummulative sums, etc.