[Edserplo_users] Update: Tidal Analysis

[Steve Loch]

Finally I have debugged the code for the Gauss Pivotal matrix inversion. We seem to be approximating the Fortran to about 6 sig. fig. We could do better by paying more attention to the inputs where there are discrepancies. Duplicating results to high precision gives confidence that we have a faithful replication of the Fortran functionality in Java.  

The Gauss Seidel inversion - an iterative process - has yet to be checked out, so all inversion is done by GP rather than GS. With computing power the way it is this hardly matters.

My attention is now focussing on the Analysis page. The font is AWFUL and I hope to have a go at changing it as a top priority. The list of constituents is all known constituents - not ideal - so it may be Samrat found it hard to do otherwise. Storage into Oracle doesn't seem to work yet. Another problem I've noted so far is the failure with key pad 7/4. The page should be opened at the size required but isn't.

The remaining BIG item, as far as completing the Edserplo replication of IRIX functionality  is concerned, relates to analysis selection on the Port Timeseries page. A recomputation of tide and residuals to match the proffered analysis (possibly one computed via the OutputPage) is supposed to ensue. It is supported in the menus but otherwise not at all. This is one of the things that Samrat was allowed to ignore.

-----------------------

One point that people may be interested in is that having converted TIRA to Java it can be called from Matlab. Coding TIRA directly in Matlab has, I think, been tried in POL but found to be extremely slow (too many loops, no doubt!). People who  want to know how to do this should contact me. 

Steve