Why this type-coercion dot at WRITE FILE?

[Ben]

Coercion dot yes!

Buffer NO!

Look for the small red circle I used to highlight the new buffer allocation.

This is NOT the case with the odd construct.

Ben

[CoastalMaineBird]

What about use flatten the typeref to string first and write the string to file instead?

Well, then you are guaranteed to make a copy of the data. That's exactly what I'm trying to avoid.

[CoastalMaineBird]

Look for the small red circle I used to highlight the new buffer allocation.

I used LV 7.1 on the Mac to open your code. I used the SHOW BUFFER ALLOCATIONS tool and I see a buffer allocation at each WRITE FILE function, whether or not you use the weird typecast. It's little black squares that flash on the function icons, not a red circle.

In other words, the typecast does nothing at all to affect the code. (That's what I would expect.)

[Chienwei]

Hi there,

Yes I'm sorry I went too far and forgot about the basics...

If your requirement is to write a datalog file without duplicating data, then I guess you can,

1. contact NI and ask them to fix their VIs (they should fix the VIs anyway);
2. wire DATALOG TYPE with a constant and try not to let the two troubles trouble you.

Afterall I don't see a perfect solution right now that would meet all your needs, other than getting a fix from NI that is.

Dan

[Ben]

Unless this acts different on a MAC?

Here is the screen shot again.

I do not get the buffer allocation with the strange construct.

You do?

Baffled,

Ben

[Ben]

I forgot the screen shot!

[CoastalMaineBird]

I verified again. With the weird typecast (which ought to do absolutely nothing, right?) wired into the DATALOG TYPE of the NEW FILE function, then BOTH of the WRITE FILE functions have a new buffer allocation (little black square).

This is LV 7.1 / OS X

It's possible that the Windows version says that it only needs to allocate ONCE, since there's not really any difference, and the Mac version is not that smart yet. I don't know.

[Chienwei]

Hi Ben,

I'm running LabVIEW on Windows 2k and I get the coericion dot and buffer allocation on file write VI.

CoastalMaineBird has mentioned that there is a miswire in your VI. The boolean should be wired to DATALOG TYPE instead of the OVERWRITE (F) input.

Dan

[Ben]

Thank you!

I missed that.

Please disregard all my previous comments.

Yes we have a buffer being created.

sigh,

Ben

[CoastalMaineBird]

Yes we have a buffer being created.

Alright, we're back at square one. The weird typecast does nothing (which is a relief in itself).

The question remains, why does a TYPEDEF incur a buffer-creation / type-coercion penalty, where an equivalent non-TYPEDEF structure does not?