Hello bercow,

The first option (using memory addresses directly in the program) I would not recommend unless your program is very, very simple. As the program grows, it will be very difficult to work directly with memory addresses.

Creating a variable for each I/O and putting them in a single data block would improve code readability and facilitate debugging compared to the previous option. However, I recommend you to use more DBs to to better organize the code.

In the third approach the program calls the Function Blocks to interact with the I/O, rather than directly referencing the physical addresses. This is a good option when you need to perform the same actions in diffenrent parts of your code. Also, it improves code understanding and makes debugging easier.  

Finally, UDTs (User Defined Types) are useful to group related input and output signals. They also improve code reusability, understanding and debugging. 

The choice depends on the requirements of the project, how complex is it and the programmer personal preferences. Usually,  the more modular and abstracted the I/O handling is, the easier it is to maintain and modify the program in the long run.

Check:Basics of PLC data types (UDT, Function blocks (FB) and Basic principles for programming of data blocks

As already stated, there are valid reasons for any specific approach.

For example, if you want to have a lamp test feature, it is best to put the outputs in a DB (or M-Tag if you must).  In that way, you can drive the dig output for a lamp directly from the PLC logic that controls the DB tag for a given output, with a parallel input NO with lamp test input.

"Tags".DigOut1                             DigOut1
---| |------------------------------------+--( )---
                                          |
"Tags".LampTest                           |
---| |------------------------------------+

bercow

Hello.

What do you do with input/output in the program? I mean how do you treat them in the code, because I've seen quite a few solutions.

- directly use I0.0 in the ladder

- put everything to DB in one function, and then using references to DB in the code

- put I/O in FB which does a few operations with them, and then we get rewriting, negations, edges, timer. A few such basic functions. And using it in LAD

- entering everything in udt and creating some logical groups.

And maybe something else.