Microsoft announces Python formulas in Excel... which have to get sent to the cloud

bedrooms, 10 months ago

I can understand the cloud part. they wanted it work on the web and phones. They do know many businesses don't want cloud, so I see a good chance they'll ship it with embedded Python eventually.

Aatube, 10 months ago

Phones can also run Python and web is already a pretty separate version, I don't see why they can't only make the web version cloud.

HidingCat, 10 months ago

Money, what else? Office 365 is a priority and this is an attempt to hasten adoption.

bedrooms, 10 months ago

At least iPhone apps usually redirect Python tasks to their servers. That's one reason there are projects like Tensorflow Light.

Aatube, 10 months ago

For instance? All the python apps I've downloaded so far seem to function offline.

conciselyverbose, 10 months ago

You can even get third party libraries, though it's limited compared to less restrictive environments.

bedrooms, 10 months ago

You can look up Tensorflow Light.

Aatube, 10 months ago

That's just machine learning which is very resource consuming. It has no relation to your purported case of phones redirecting all python tasks to servers.

bedrooms, 10 months ago

Okay, I dug more to find out I'm wrong.

But isn't ML technology a thing Excel offers with its new Python interface?

Aatube, 10 months ago

Not really. I don't expect them to have a cloud instance running that long.

Aatube, 10 months ago

Someone on lemmy.world pointed out the FOSS xlwings also exists.

Limitless_screaming, 10 months ago

Are they not gonna give a bullshit reason for this? Just straight up give us your data and it's secure on our Azure instances?

kamills, 10 months ago

Seems a bit odd to not have the option to run python locally on my own machine

style99, 10 months ago

It would be very strange if you couldn't just export your data to csv and then run a python script in a shell on it.

flamingo_pinyata, 10 months ago

Because it was so much easier to send data to the cloud than embed a Python interpreter. 🤦

I wouldn’t be surprised if there already is one in the monstrosity that is Excel

lemonflavoured, 10 months ago

My guess is that they are seeing this as less likely to become a security hole.

earthling, 10 months ago

Yep. Everyone in the thread asking this question seems clueless to me. Macros are already a threat. I can’t imagine what a shitshow full on python would be.

Aatube, 10 months ago

The Python API they gave doesn't have disk access. Maybe somebody'll discover an exploit but that's for everything.

nottheengineer, 10 months ago

Python is slow enough as is, who the fuck thought adding a web request to that was a good idea‽

sugar_in_your_tea, 10 months ago

Python is fast enough, and generally quite a bit faster than complex spreadsheet formulas. I’ve seen formulas that can take minutes that Python could do in seconds. A web request isn’t going to matter that much.

Aatube, 10 months ago

Web requests are <em>very</em> slow compared to CPU computations, not to mention that time has to be doubled since it's a two-way route.

sugar_in_your_tea, 10 months ago

Sure, but it happens once. So as long as Python saves you more than the half second or so round trip, it’ll be preferable.

I’d prefer it to be embedded, but I can absolutely see it being useful even if it’s cloud-only.

Aatube, 10 months ago

Wouldn't it need to send to the cloud and back every time you change one of the cells the script is depending on?

sugar_in_your_tea, 10 months ago

Yup, most likely. I’m saying the total calculation w/ Excel formulas is often greater than that round trip + computation time w/ Python. Excel formulas are pretty slow.

Aatube, 10 months ago

I haven't seen how they were slower than just python.

sugar_in_your_tea, 10 months ago

Think huge formulas. I’ve seen formulas take minutes in Excel, but seconds when implemented better in a script. If you haven’t worked with massive formulas, you’re not the target market here.

Another application is accessing external APIs that don’t have internal support. If you’re accessing an external API, you’re already paying a network overhead cost, so adding another isn’t going to matter much.