Question

What exactly is the difference between bytecode and a parse tree, specifically the one used by Perl? Do they actually refer to the same concept, or is there a distinction?

I’m familiar with the concept of bytecode from Python and Java, but when reading about Perl, I’ve learned that it supposedly executes a parse tree (instead of bytecode) in its interpreter.

If there actually is a distinction, what are the reasons for Perl not using bytecode (or Python not using parse trees)? Is it mainly historical, or are there differences between the languages that necessitate a different compilation/execution model? Could Perl (with reasonable effort and execution performance) be implemented by using a bytecode interpreter?

Answer 1

What Perl uses is not a parse tree, at least not how Wikipedia defines it. It’s an opcode tree.

>perl -MO=Concise -E"for (1..10) { say $i }"
g  <@> leave[1 ref] vKP/REFC ->(end)
1     <0> enter ->2
2     <;> nextstate(main 49 -e:1) v:%,{,2048 ->3
f     <2> leaveloop vK/2 ->g
7        <{> enteriter(next->c last->f redo->8) lKS/8 ->d
-           <0> ex-pushmark s ->3
-           <1> ex-list lK ->6
3              <0> pushmark s ->4
4              <$> const[IV 1] s ->5
5              <$> const[IV 10] s ->6
6           <#> gv[*_] s ->7
-        <1> null vK/1 ->f
e           <|> and(other->8) vK/1 ->f
d              <0> iter s ->e
-              <@> lineseq vK ->-
8                 <;> nextstate(main 47 -e:1) v:%,2048 ->9
b                 <@> say vK ->c
9                    <0> pushmark s ->a
-                    <1> ex-rv2sv sK/1 ->b
a                       <#> gvsv[*i] s ->b
c                 <0> unstack v ->d
-e syntax OK

Except, despite being called a tree, it’s not really a tree. Notice the arrows? It’s because it’s actually a list-like graph of opcodes (like any other executable).

>perl -MO=Concise,-exec -E"for (1..10) { say $i }"
1  <0> enter
2  <;> nextstate(main 49 -e:1) v:%,{,2048
3  <0> pushmark s
4  <$> const[IV 1] s
5  <$> const[IV 10] s
6  <#> gv[*_] s
7  <{> enteriter(next->c last->f redo->8) lKS/8
d  <0> iter s
e  <|> and(other->8) vK/1
8      <;> nextstate(main 47 -e:1) v:%,2048
9      <0> pushmark s
a      <#> gvsv[*i] s
b      <@> say vK
c      <0> unstack v
           goto d
f  <2> leaveloop vK/2
g  <@> leave[1 ref] vKP/REFC
-e syntax OK

The difference between Perl’s opcodes and Java’s bytecodes is that Java’s bytecodes are designed to be serialisable (stored in a file).