CFR uses numerical literals instead of constants about cfr

[leibnitz27]

This is (partially) handled in InstanceConstants

cfr/src/org/benf/cfr/reader/bytecode/analysis/opgraph/op3rewriters/InstanceConstants.java

Line 89 in d6f6758

Map<Object, Expression> visibleInstanceConstants = ConstantLinks.getVisibleInstanceConstants(thisType, refSearchType, b, state);

However, while it's a 'nice' thing, it tends to lead to completely confusing code in all but the most obvious cases; it's been a while so I may be mis remembering, but GUI code in particular uses a lot of magic constants , so the code to do this is very pessimistic. ( while a false negative is annoying, a false positive is downright confusing ).

[wodin]

The name of the constant is not stored in the constant pool in the bytecode. The disassembled code looks like this:

     419: aload_0
     420: dload         23
     422: ldc2_w        #117                // double 1.6777216E7d
     425: dmul
     426: invokestatic  #66                 // Method org/bukkit/util/NumberConversions.floor:(D)I
     429: putfield      #119                // Field secondError:I

So it's just fetching the value stored at index 117 in the constant pool and pushing it onto the stack. There's no name associated with that value.

So the best CFR would be able to do is guess that it's the same as the gridSize constant. Is that what you're proposing?

[DecompExplorer]

The name of the constant is not stored in the constant pool in the bytecode. The disassembled code looks like this:

     419: aload_0
     420: dload         23
     422: ldc2_w        #117                // double 1.6777216E7d
     425: dmul
     426: invokestatic  #66                 // Method org/bukkit/util/NumberConversions.floor:(D)I
     429: putfield      #119                // Field secondError:I

So it's just fetching the value stored at index 117 in the constant pool and pushing it onto the stack. There's no name associated with that value.

So the best CFR would be able to do is guess that it's the same as the gridSize constant. Is that what you're proposing?

Yes, that's what I'm proposing. Thank you for clarifying that!

[DecompExplorer]

This is (partially) handled in InstanceConstants

cfr/src/org/benf/cfr/reader/bytecode/analysis/opgraph/op3rewriters/InstanceConstants.java

Line 89 in d6f6758

Map<Object, Expression> visibleInstanceConstants = ConstantLinks.getVisibleInstanceConstants(thisType, refSearchType, b, state);

However, while it's a 'nice' thing, it tends to lead to completely confusing code in all but the most obvious cases; it's been a while so I may be mis remembering, but GUI code in particular uses a lot of magic constants , so the code to do this is very pessimistic. ( while a false negative is annoying, a false positive is downright confusing ).

Thank you for your explanation.