Do you think this behaviour of built-in functions error() and warning() is odd

I find that warning('message', a1, a2,...) treats args a1 etc differently to warning('message_id', 'message', a1, a2, ..., an). The same is true for error but a surprise to me. I discovered this when I tried to show the size of an array in the warning by using size(X) as one of the arguments. What do you reckon to this behaviour, which I found after MLINT nagged me to add a message identifier but coughed when I did? Example shown below:

    >> warning('%d %d', [1 2])
Warning: 1 2 
>> warning('msg:id', '%d %d', [1 2])
??? Error using ==> warning
Formatted arguments cannot be non-scalar numeric matrices.

Answer 1

[Sean de Wolski]

It means exactly what it says:

warning('msg:id', '%d %d', 1, 2)

Use 1,2 as scalars.

I do see how this differs from sprintf() where the former was acceptable.

sprintf('%d %d', [1,2])

[Julian]

Well I agree that the warning is specific; nevertheless the behaviour is a surprise to me and is not as described in the documentation. It seems odd to me that simply following an MLINT recommendation can actually break working code!

[Sean de Wolski]

Ahh, I see what tyou're saying:

warning('%d %d', [1 2])

works, but MLINT is yelling.

warning('msg:id','%d %d', [1 2])

errors.

[Sean de Wolski]

I've submitted an enhancement request regarding this behavior.

[Julian]

I did not appreciate you were with TMW; thanks for putting in an an enhancement request - so I accept your answer. I would add that making enhancement requests more visible and open to comment would be very welcome - either via MATLAB Central or through the private support web-site. Quite a lot of discussion in Answers in this domain - and I hate to submit enhancement requests.