I'm aware of all of the failing tests. They're compilation error details, it worked on my Ubuntu machine, but it's failing here. That's normal. I'll solve them.

Concerns

* [ ]  Naming of `print_t` only member to be 'text', is that a good choice?

I think so.

* [ ]  Now, with `print` only supporting single argument of character type only. Should we throw an error when we encounter a `text` member that isn't of type `character` or should we only assert or to not even care about that?

The verify pass should have a check for this.

* [ ]  Is removing `end` and `sep` from `print` gonna affect lPython?

Yes. LPython will need to use StringFormat also. We can write a utility function that does it automatically for LPython.

* [ ]  Should We use `make_print_t_util()` that creates a print node with stringformat wrapped in it if necessary, or should build such a thing on our own each time we want print bunch of arguments.

I think so.

* [ ]  `replace_symbolic` pass visits `print` and sees if it has some ready compile time values (In the old `print` node definition, not the one in this PR), then it replaces the argument with value (as I understand) ; Do we benefit from this visitor?, if so we would need to add `values` member into `print` structure.

Print is a statement, so it doesn't have an evaluation. But StringFormat is an expr so it should have value that we will eventually fill if the formatting can be done at compile time.

The replace symbolic is for symbolic (SymPy like) expressions. The Print replacement allowed printing the expressions. Those should be now handled in StringFormat. @anutosh491 can help with this later. For now it's fine, we can even disable this pass in LFortran if needed. We should add Fortran tests for symbolics also, to ensure things work. This is all unrelated.

I'll review this by tonight.

@assem2002 let's finish this, I think it is almost done.

It should be ready now.

Thank you. I think this is great! Can you please resolve the conflicts?

To summarize, this PR makes the Print node accept just the text as a required argument. Much simpler!

@Pranavchiku, @gxyd can you please also review this PR if you see any issues?

Looks good. I shall address this TODO commit here (3782c7a) once this goes in

We need to print the empty space for lfortran --print-leading-space examples/expr2.f90 (one way or the other), then we can merge.

--print-leading-space works as intended. examples/expr2.f90 works fine with flag --print-leading-space.
The resulting ASR of lfortran examples/expr2.f90 --print-leading-space is now like this :

          (FileWrite
              0
              ()
              ()
              ()
              ()
              [(StringConstant
                  " "
                  (Character 1 1 () PointerString)
              )]
              ()
              (StringConstant
                  ""
                  (Character 1 0 () PointerString)
              )
              ()
          )
          (Print
              (StringFormat
                  ()
                  [(Var 2 x)]
                  FormatFortran
                  (Character -1 0 () PointerString)
                  ()
              )
          )]
      )

With the suggestion you made here #4696 (comment) about adding just an arg into the stringFormat node instead of creating a whole write statment node, We can transfer the resulting ASR to look something like this :

(Print
            (StringFormat
                ()
                [(StringConstant
                    " "
                    (Character 1 1 () PointerString)
                )
                (Var 2 x)]
                FormatFortran
                (Character -1 0 () PointerString)
                ()
            )
        )]
    )

but the final print would be --> ' ' + four-default-spacing(set by the backend) + var x, so spacing won't be just 1 single empty space it would be 5 empty spaces.

Let me know what you think is better to use.