Fix comments in REPL - fixes #78

[ncw]

Before this change, entering a comment in the REPL caused the REPL to
read the comment indefinitely effectively breaking it.

After this change the behaviour should be exactly the same as python3/

[ncw]

Can you review this for me @corona10 ?

[codecov-io]

Codecov Report

Merging #79 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@ Coverage Diff @@## master #79 +/- ##
==========================================
+ Coverage 68.65% 68.66% +<.01% 
==========================================
Files 59 59 Lines 10525 10528 +3 ==========================================
+ Hits 7226 7229 +3 
Misses 2790 2790 Partials 509 509

Impacted Files	Coverage Δ
repl/repl.go	100% <100%> (ø)	⬆️
symtable/symtable.go	94.23% <0%> (ø)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update eb115a9...9a4df56. Read the comment docs.

[corona10]

This patch will fix this issue #78 .
But the input of '#' should be handled as same as below code?
Should we update grammar file for this?

[Gpythondev]
-os/arch: darwin/amd64-goversion: go1.10>>>defa(): pass>>>

[corona10]

This is the example of pypy and rustpython.

Python3.6.1 (dab365a465140aa79a5f3ba4db784c4af4d5c195, Feb182019, 10:53:27)
[PyPy7.0.0-alpha0withGCC4.2.1CompatibleAppleLLVM10.0.0 (clang-1000.11.45.5)] ondarwinType"help", "copyright", "credits"or"license"formoreinformation.
Andnowforsomethingcompletelydifferent: ``"it'slikelytemporaryuntilforever" arigo''>>>>#>>>>#>>>>

WelcometothemagnificentRustPython0.1.0interpreter 😱 🖖
>>>>>#>>>>>

[corona10]

And it still doesn't work well on my local laptop with branch fix-78-replbcb3785

Python 3.4.0 (none, unknown)
[Gpython dev]
- os/arch: darwin/amd64
- go version: go1.10
>>> #
... #
...

[ncw]

And it still doesn't work well on my local laptop with branch fix-78-replbcb3785

Python 3.4.0 (none, unknown)
[Gpython dev]
- os/arch: darwin/amd64
- go version: go1.10
>>> #
... #
...

Ah you are right this is broken :-(

The pypy and rustpython ways of doing it are

• a whole lot more logical in my opinion
• much easier that emulating python3 exactly

Shall I make it work like that?

[corona10]

@ncw
Yes, please.
IMHO, we don't have to follow everything of CPython :)

[ncw]

OK I did that! PTAL

[corona10]

@ncw

gpython/parser/lexer.go

Line 375 in 2609237

func (x*yyLex) Lex(yylval*yySymType) (retint) {

Can we fix this issue on lexer level approach?

[corona10]

@ncw
This PR is pending for 2 weeks.
Should we merge this PR for this time and later fix it?
Or do you have any ideas?
Thanks!

[ncw]

I think this is good to go now.

I don't think fixing it in the lexer is the right place - this is specifically to do with how the REPL works and in particular how the single input grammar works which is a bit strange but it is how the Python grammar is defined.

I'll merge this now and if there are any problem with it I'm sure we can fix them up :-)