Refactor entire thing into Matlab class?

[micahjsmith]

Some inefficiencies that stand out include:

• needing to create a new GlobalOptions for every call
• needing (further) to load the API key from the text file for every call
• needing to call parseVarargin for many calls, whereas parworkers, pseudo, and units presumably should all be global options - that is, they are set once at the beginning of the session and the explicitly changed as desired

These behaviors can be achieved easily by having an object of class fred or of class FredFetch be created at the beginning of the session:

f = fred(MY_API_KEY,'defaultUnits','lin')
% alternately
f = FredFetch(MY_API_KEY)
set(f, 'defaultParWorkers',3) % one way to set/get fields of object
f.defaultOtherOption = 'hello' % another way to set/get fields of object

Then, a future call is made as:

latest(f, 'GDCP1')
% or, completely equivalent syntactically
f.latest('GDPC1')

The object f already has "embedded" all the relevant global options and the API key without needed to do everything from scratch. This is also what the actual Matlab Datafeed Toolbox fred routines (as for all other data sources) do.

See my bls interface for a simple example of what this looks like/how it is implemented.

[micahjsmith]

Additional advantage of setting global options in a first, initial step: f.defaultParWorkers can be specified, but then the class tests for the existance of a Parallel Computing Toolbox license and sets to 0 otherwise.

[MattCocci]

Into it. But I want to implement time-aggregation of series first. I'd rather devote coding effort to that at this point. That's really the last major feature that's missing.

Then longer-term, I think this is a good idea.

[micahjsmith]

Seems fair enough