Re: [Openexr-devel] KeyKodes (a recap)

[Ken McGaugh]

Florian Kainz wrote:
> I don't think IlmImf should include hardwired lists of film
> manufacturers and their film types, but some of this information
> is necessary to build a human-readable string from the numeric
> keycode.
>

Fair enough.

So it seems we finally have a plan.  Basically, all we are going to do
is add a perfsPerFrame int member to the KeyCode class.  If unknown,
the value of this will be 0 and the interpretation of that will be
left to the software reading the exr.

As for the perfOffset, we will interpret it as the number of perforations
between the reference mark the *last* perf in the frame.  Using integer
math, we can compute the frame and alignment as follows:

    frameOffset = perfOffset / perfsPerFrame
    frameAlignment = perfsPerFrame - (perfOffset % perfsPerFrame)

where frameOffset is the "+ 00" bit in the human readable keycode,
and frameAlignment is the "p1" at the end.

Often the alignment information will not be available and the perfOffset
will always be a multiple of the perfsPerFrame.

We are also going to add the standard string attributes "filmFormat"
and "slateInfo".

I'm going to be bold and start using this today as I have scans to convert.

Thanks again to everybody and I apologise to those who are bored stiff of this
conversation.  Now we need to discuss timecode, which is just an unsigned 32-bit
integer, but IlmImf only supports signed 32-bit integers.  What to do... :)

--Ken