How to Make a Huge Struct Accessible to Many Functions?

Question

Rightia Rollmann 2017-3-17

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/330391-how-to-make-a-huge-struct-accessible-to-many-functions

回答： Image Analyst 2017-3-17

I have a huge struct loaded by a function. I have other functions that will use some of the fields of the struct to calculate something. What is the best way to make this struct accessible to all other functions? More info: this struct will only be read and no information by other functions will be overwritten on it.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Adam 2017-3-17

I'm not really sure how this question differs from this one where you appeared to be asking the same thing and Steven Lord gave you a link that explains about sharing between callbacks.

The answer in general is still the same, though with specific code, which you haven't included, may vary.

The fact that you have a 'huge struct' needed by many functions in the first place sounds like the start of the problem. Functions should ideally be given no more than they need to perform their own task within the overall system.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jan 2017-3-17

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/330391-how-to-make-a-huge-struct-accessible-to-many-functions#answer_259197

Avoid globals, because they impede the debugging and expanding of code.

If y struct is provided as input, only the struct pointer is copied, but Matlab does not duplicate the contents of the struct. Therefore it does not matter if the struct is huge or tiny. Forwarding the struct as an input will work efficiently. As long as you do not reply the struct as output, there is no danger of changing its contents "by accident".

But Adam's objection is correct in general: Providing all information to all subfunctions seems to be a bad design. If the code is large and written by multiple programmers, a modification of this struct will be dangerous, because it is hard to check, which functions are effected. The less data are forwarded, the less side-effects must be checked for modifications. Or in other words: The leaner the code, the less chances to fail.

Phew, don't ask me for a proof :-)