On 8/19/20 3:08 PM, David G. Johnston wrote: > On Wed, Aug 19, 2020 at 11:51 AM Pól Ua Laoínecháin <linehanp@xxxxxx > <mailto:linehanp@xxxxxx>> wrote: > > > I think my *MAJOR* problem is that I've developed what is, > essentially, a totally brute force approach - and this simply won't > work at the scenario becomes more complex - take a look at the CASE > statement - it's horrible and would only become exponentially worse as > the number NULLs rises. > > So, my question is: Is there a recognised technique (using SQL only, > not PL/pgSQL - soutions based on the latter are easy to find) whereby > I can do a basic Linear Interpolation? > > I don't have a recognized technique, nor care to ponder one right now, but what > you've described would best be done in pure SQL using WITH RECURSIVE, which > provides an iterative approach to SQL result building. Which is more commonly > done in a procedural language. The algorithm you describe is an iterative > algorithm and so I'm wondering why place the arbitrary restriction on using pure > SQL when it likely wouldn't provide a very readable nor performant solution > relative to a procedural (pl/pgsql or otherwise) one? Might want to consider using PL/R with something like the R imputeTS package: https://cran.r-project.org/web/packages/imputeTS/readme/README.html HTH, Joe -- Crunchy Data - http://crunchydata.com PostgreSQL Support for Secure Enterprises Consulting, Training, & Open Source Development
Attachment:
signature.asc
Description: OpenPGP digital signature
