It isn't even that hard to get a basic implementation of such functionality. Not sure why this is missing.
Just make a directed graph out of cell references. Get the strongly connected components. All of the components with more than one vertex contain a loop and can't be computed. Remove those and you get a directed acyclic graph. Now you can topological sort them for cell dependencies, and start calculating them in the order they are needed.
Or if your language has a dataflow implementation it is even easier. Haven't seen one of these for Obj-C though.
It may be easy, but it'll still take time to develop. If it's not critical and won't keep you from getting your initial batch of users, it's much better to launch and add it in 3 weeks, rather than waiting an extra 3 weeks to launch.
I may have sounded a bit negative, but I don't think it was a "bad" move, especially seeing as how I am personally excited about Grid even without it currently having this feature.
It just shouldn't come as a surprise when a lot of people question the choice of leaving out the #1 feature of spreadsheets. Even on version one.
You forgot parsing the formulae. You forgot having to decide which functions to include and implementing them. You forgot testing. Sure those aren't super hard, but minutes add up to hours and hours add up to days which turn into weeks. Most importantly, you forgot the UI. Building the UI for entering formulae - one that makes sense on the iPad and not just a copy of the one from Excel or Numbers (those suck) - would take a while.
Apple could've thrown copy and paste into the first iPhone, but they didn't because they didn't feel they had it right.
They might be in the process of actually designing an interface for formulas that actually expands the reach of such functionality. You know, something that the "I'm bad at math" crowd could actually dig into, make use of, and learn within.
If you release a basic not-that-hard-to-implement version of the functionality, it's WAY harder to drop in a new, innovative model afterwards. I think there's a lot of sense in waiting until they can get it right, if they're planning on doing something different with it--which, judging by the other UI departures they've made, I wouldn't be surprised if they are.
Just make a directed graph out of cell references. Get the strongly connected components. All of the components with more than one vertex contain a loop and can't be computed. Remove those and you get a directed acyclic graph. Now you can topological sort them for cell dependencies, and start calculating them in the order they are needed.
Or if your language has a dataflow implementation it is even easier. Haven't seen one of these for Obj-C though.
I will add that this looks pretty cool, though.