Extending support for Matlab

From OPeNDAP Documentation
Revision as of 23:12, 9 March 2010 by Jimg (talk | contribs) (New page: Expanding OPeNDAP’s Support for Matlab James Gallagher 17 April 2009 Why does this matter? OPeNDAP’s fundamental mission is to provide support for the Data Access Protocol and to fur...)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
⧼opendap2-jumptonavigation⧽

Expanding OPeNDAP’s Support for Matlab James Gallagher 17 April 2009

Why does this matter?

OPeNDAP’s fundamental mission is to provide support for the Data Access Protocol and to further its extension by keeping its features in sync with changes taking place in the rapidly evolving sphere of networked data systems. One important change that has taking place in the last decade is the rise in prominence of data analysis tools that effectively replace programming languages like FORTRAN and (to some extent) C. Two such examples are the Matlab and IDL tools developed by Mathworks, Inc. and RSI, Inc., respectively. Many users of the DAP depend on these tools day in and day out, so continued support for them is crucial both for those users and for the viability of the DAP.

What is the problem?

Unfortunately our support for these all-important tools has lagged behind our efforts in other areas. Throughout its existence, OPeNDAP has maintained an excellent program of server upgrades, and that software is well regarded by many who rely on it to serve their data. Furthermore, even though the project is yet to formally complete, the NSF SDCI effort to modernize our netCDF client library has been, so far, an outstanding success. Contrast these efforts with the piecemeal approach taken towards supporting command extensions for Matlab so data from DAP servers can be read directly into this analysis tool. Because support has been sporadic, the command extension often lags behind Matlab by as much as a whole year and does not track changes in the APIs it supports. Compounding the problems created by this situation is that the current design and implementation of the command extension is cumbersome and expensive to maintain. A bad situation keeps getting worse because the effort needed to sustain the extension is so costly.

Our proposed solution

We propose to redesign and reimplement the Matlab command extensions using our new C library for client access (called OC) that has been developed as part of the NSF SDCI project to improve our netCDF client-library software. There are advantages to switching to this new library for this command extension. For the Matlab client, a redesign will eliminate a chief complaint of users that the two programs that make up the command extension are too hard to install and use. Adopting a single library as the base for both these and the netCDF client-library will allow us to amortize many support costs across three clients (our IDL client already is being rewritten to use the new library), creating a clearer path for reliable support into the future.