Re: rewrite of structs - advice sought

[John W. Eaton]

On 23-Jun-2010, fork wrote:

| > May I suggest enhancing the
| > | basic load and save operation, implementing something similar to R data
| > | frame, which are:
| > |   A data frame, a matrix-like structure whose columns may be of differing
| > | types (numeric, logical, factor and character and so on).
| > | 
| > | Basically, load and save should try to return a structure array in case 
of a
| > | first non-comment line of fields name, a matrix if all values are of 
similar
| > | types, and an error code otherwise ?
| 
| (this should be a different thread if we really want to pursue it)
| 
| data.frames (and smart categorical data handling) are some of the few things I
| miss about R.  It is really nice to be able to write
| 
| data[10:15, 'colname']
| 
| and get something reasonable.
| 
| > Feel free to enter a feature request in the bug tracker, but please
| > remember that Octave is developed by volunteers and you may have to
| > provide some additional incentive to get someone interested in working
| > on a particular project.
| 
| If there were a well designed data.frame-ish idea, I think it would be fairly
| easy (as in 100 lines or so) to write an m-function to handle the OP's basic
| idea for importing.  We have scanf and split functions...
| 
| > Also, we have compatibility with Matlab (not R) to consider.  Perhaps
| > it would be best if someone were to write a function for Octave that
| > is compatible with Matlab's importdata function.
| 
| While this would be cool, it wouldn't address the issue of different column
| types or column names.  Unless I am misreading the online docs.  

It looks to me like spreadsheet data doesn't have to be all numeric,
but I have no experience with actually using the function, so maybe I
misunderstand what it can do.

| And I personally think sophisticated data-handling would be a good reason to
| allow non-TMW based functions and protocols.  I am all for Octave being a 
strict
| superset of matlab, and in this case there is no good matlab approach.

There is nothing preventing you from writing and contributing a set of
functions for reading R-like data frames (even R data files) into some
kind of struct- or cell-array data structure with all the appropriate
tags and data included in one object.  But I think that overloading
Octave's load function for that purpose is not the right thing to do.

jwe