> FFS, there is a conference on how to not screw things up in real life with Excel.
> In Excel, I change one or two cells and see everything update immediately
This is the problem - it's all too easy to change cells, misclick, and generally make mistakes. When you do things from a programming language, you need to be explicit about operations, and explicit about what changes you're making.
Spreadsheets should be for entering data only, R is for processing data.
All the examples of enter data, create a chart, oh let's change/update some data, oh let's add something to the file, etc..., are why you end up with incredibly convulated spreadsheets that inevitably are filled with errors.
Enter your data with whatever tool you want (spreadsheet works for this), then process it (R works for this), then shove it in a database. Nice and easy, not error prone, and every tool can have easy access since databases are ubiquitous.
The user actions that lead to mistakes in Excel are the same with any other environment: typing formulas. If I'm doing an engineering or financial calculation and misplace a decimal point, flip a sign, or forget an ABS(), I'm going to get the same wrong result in R or Excel. The upside of Excel is that the feedback is immediate so you have a chance of spotting the error if you're expecting a certain output. There is no clean solution to that regardless of environment. It's just as easy to make mistakes when you're changing code or changing a spreadsheet.
R has the advantage of being easier to do a proper diff between files if a change was made in error, but most people don't think they made a change in error. The change was intentional; they just got it wrong.
> The upside of Excel is that the feedback is immediate so you have a chance of spotting the error if you're expecting a certain output.
Only some errors are obvious just from a glance at the result set. For other errors, Excel makes the developer's job much harder. Subtle bugs can hide the code of an individual cell among thousands, and the user has no means of clearly abstracting that code out.
I completely agree. There are some things you can do in Excel that are nearly impossible to track down. There are auditing tools that help, but you can truly ruin a spreadsheet in undetectable ways if you want (or more likely, don't want).
> In Excel, I change one or two cells and see everything update immediately
This is the problem - it's all too easy to change cells, misclick, and generally make mistakes. When you do things from a programming language, you need to be explicit about operations, and explicit about what changes you're making.
Spreadsheets should be for entering data only, R is for processing data.
All the examples of enter data, create a chart, oh let's change/update some data, oh let's add something to the file, etc..., are why you end up with incredibly convulated spreadsheets that inevitably are filled with errors.
Enter your data with whatever tool you want (spreadsheet works for this), then process it (R works for this), then shove it in a database. Nice and easy, not error prone, and every tool can have easy access since databases are ubiquitous.