Add to_buffer(string) and use it when initializing a buffer

[Veracity0]

I've wanted to be able to do this for a long time:

buffer b = "Initial contents";

Now I can.

I also moved creation of the (buffer) return value of visit_url() to the top of the method, rather than duplicating the creation in two places.

[codecov]

Codecov Report

Merging #1907 (d8cf23b) into main (68b9dca) will increase coverage by 0.00%.
The diff coverage is 82.85%.

Additional details and impacted files

@@            Coverage Diff            @@
##               main    #1907   +/-   ##
=========================================
  Coverage     36.37%   36.38%           
- Complexity    18830    18833    +3     
=========================================
  Files          1083     1083           
  Lines        166605   166619   +14     
  Branches      35391    35393    +2     
=========================================
+ Hits          60609    60622   +13     
+ Misses        96110    96109    -1     
- Partials       9886     9888    +2     

Files Changed	Coverage Δ
...et/sourceforge/kolmafia/textui/RuntimeLibrary.java	40.55% <69.23%> (+0.09%)	⬆️
src/net/sourceforge/kolmafia/textui/Parser.java	94.14% <90.90%> (-0.07%)	⬇️

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 68b9dca...d8cf23b. Read the comment docs.

[midgleyc]

I think this is good to have in main, but I think there are a fair few scripts (e.g. Ezandora's) which define

buffer to_buffer(string str)
{
  buffer result;
  result.append(str);
  return result;
}

for the convenience, and which will now conflict and error. I think s.replace_string("", "") is also popular.

[Veracity0]

Yeah. Ezandora's scripts concern me.
I made an announcement about this, since there will be much lamentation.
Perhaps it will inspire her to upgrade her scripts. :)

On the other hand, we could make the compile error into a warning. If the script is not depending on the RuntimeLibrary version, no big deal if it redefines it - even to something not equivalent.

[Veracity0]

And, oh - for replace_string, we don't want a string to auto-coerce into a buffer.

Operator is allowed to check coercion based on context: "assign", "return", "parameter".
But that is checked after auto-coercion - finding a function to do it for you - is done.

Hmm.

[BadHorseMonkey]

I think this is good to have in main, but I think there are a fair few scripts (e.g. Ezandora's) which define
...

Also, your Bastille script. And TourGuide. And those are just the ones I have installed.

[Veracity0]

Interesting.

    if (f.overridesLibraryFunction()) {
      String buffer = "Function '" + f.getSignature() + "' overrides a library function.";
      this.warning(f.getLocation(), buffer);
    }

works nicely if I declare the library function like this:

    params = new Type[] {DataTypes.STRING_TYPE};
    functions.add(new LibraryFunction("to_buffer", DataTypes.BUFFER_TYPE, params));

but not if I declare it like this:

    params = new Type[] {DataTypes.STRICT_STRING_TYPE};
    functions.add(new LibraryFunction("to_buffer", DataTypes.BUFFER_TYPE, params));

There is no way for the user to declare a function whose argument is a STRICT_STRING_TYPE

[Veracity0]

I made this script:

// The following should generate a warning
buffer to_buffer(string initial)
{
    print("initial = '" + initial + "'");
    buffer buf;
    buf.append(initial);
    return buf;
}

buffer buf1 = to_buffer("Initial content 1");
print("buf1 content = '" + buf1.to_string() + "'");

buffer buf2 = "Initial content 2";
print("buf2 content = '" + buf2.to_string() + "'");

buffer buf3 = $path[Heavy Rains];
print("buf3 content = '" + buf3.to_string() + "'");

which yields:

Function 'to_buffer(string)' overrides a library function. (to_buffer_test.ash, line 2, char 8 to char 33)
initial = 'Initial content 1'
buf1 content = 'Initial content 1'
initial = 'Initial content 2'
buf2 content = 'Initial content 2'
initial = 'Heavy Rains'
buf3 content = 'Heavy Rains'

The good news is that the user defined function really does override the library function.
The bad news is that I didn't REALLY want the $path[] to get coerced into a string.
Which is to say, I'd prefer STRICT_STRING_TYPE.

I don't see an immediate clean solution to the STRING_TYPE != STRICT_STRING_TYPE issue.
Function.paramsMatch:

    while (it1.hasNext() && it2.hasNext()) {
      VariableReference val1 = it1.next();
      VariableReference val2 = it2.next();
      Type paramType1 = base ? val1.getType() : val1.getRawType();
      Type paramType2 = base ? val2.getType() : val2.getRawType();
      if (!paramType1.equals(paramType2)) {
        return false;
      }

It compares Type using equals(). No special casing to pretend STRING_TYPE is the same as STRICT_STRING_TYPE.
Until/unless we rethink how STRICT_STRING_TYPE is implemented, I have two options:

1. to_buffer(string) is declared with a STRING_TYPE argument and we get useful override warnings
2. to_buffer(string) is declared with a STRICT_STRING_TYPE argument and we get no warning.

I'm inclined to do #1, for now.