consider the following example:

[MATH]j:2[/MATH]

[MATH lang=ENG]for(j,range(1,4),a:j)[/MATH]

[MATH]j=4[/MATH]

Now edit the loop expression and change it to any expression without assignment operator with or without j in it

[MATH lang=ENG]for(j,range(1,4),a)[/MATH]

When leaving the math region, nothing happens.

[MATH]j=4[/MATH]

As soon as you trigger re-calculation manually, j changes to 2, indicating that the loop iteration is not performed in the global namespace.

[MATH]j=2[/MATH]

Still there is iteration as you can see using [MATH]trace(j)[/MATH] as loop body.

[MATH lang=ENG]for(j,range(1,4),trace(j))[/MATH]

I think, that changes in expressions should trigger re-calculation if that option is set in the options.

Also, I really discourage the use of implicit variable scoping magics. I guess that it is not by intent to overwrite the global imaginary unit i depending on what precisely you do inside iterations with i as counter.

If the mentioned dependency is a feature and not a bug I would like to learn how this contributes to comprehensable documents.

Best regards, Martin Kraska

Without reading this post, I just posted the same thing in the clear discussion.