Strange behavior due to missing semicolon

[yoshinari-nomura]

Hi, I started working with tree-sitter-openscad yesterday.
I noticed some strange behavior.

Create the test as follows:

cargo install tree-sitter-cli
git clone git@github.com:openscad/tree-sitter-openscad.git
cd tree-sitter-openscad
echo 'test("AB")' > test-ab.scad
echo 'test("ABC")' > test-abc.scad

tree-sitter parse test-ab.scad shows:

(source_file [0, 0] - [1, 0]
  (ERROR [0, 0] - [0, 10]
    (module_call [0, 0] - [0, 10]
      name: (identifier [0, 0] - [0, 4])
      arguments: (arguments [0, 4] - [0, 10]
        (string [0, 5] - [0, 9])))))
test-ab.scad	Parse:    0.14 ms	    77 bytes/ms	(ERROR [0, 0] - [0, 10])

tree-sitter parse test-abc.scad shows:

(source_file [0, 0] - [1, 0]
  (transform_chain [0, 0] - [0, 11]
    (module_call [0, 0] - [0, 11]
      name: (identifier [0, 0] - [0, 4])
      arguments: (arguments [0, 4] - [0, 11]
        (string [0, 5] - [0, 10])))))
test-abc.scad	Parse:    0.12 ms	    98 bytes/ms	(MISSING ";" [0, 11] - [0, 11])

Why are the two different?
I don't know which behavior should be desired.

[mkatychev]

This seems very odd, once I'm at my computer I will take a look

[mkatychev]

@yoshinari-nomura: apologies for the late reply, I am indeed able to reproduce the error.

[mkatychev]

So it seems the expected behaviour should be to have a statement/semicolon at the end:

tree-sitter-openscad/grammar.js

Lines 254 to 258 in 09ed147

	transform_chain: $ => seq(
	repeat($.modifier),
	$.module_call,
	$.statement,
	),

As you can see in your abc, the last line says (MISSING ";" [0, 11] - [0, 11]) which is what should be reported for both 🤔

EDIT: It seems that the classifier may be expecting a minimum length of the source_file to have (ERROR) vs. (MISSING) propagate: t("ABA") produces error and _test("AB") produces missing

[yoshinari-nomura]

@mkatychev, thanks for your great inspection.

EDIT: It seems that the classifier may be expecting a minimum length of the source_file to have (ERROR) vs. (MISSING) propagate: t("ABA") produces error and _test("AB") produces missing

Interesting. Turns out I was lucky to choose "test" as the name 😄

I am developing an indentation system for Emacs using tree-sitter-openscad:
https://github.com/yoshinari-nomura/scad-ts-mode
So, I did such a trivial test and discovered this.

According to your assessment, it seems to be a practically harmless but,
length of the transform_chain would be matter instead of source_file.

(1) This emmits the error:

test("AB");
test("AB")

(2) But this emmits missing:

test("AB")
  test("AB")

In the handling of indentation, case (1) often seems to occur when trying to add something to existing code.

BTW, I wonder why ERROR node is shown in the tree, but MISSING is not. Can I get the MISSING from the parsed tree?

[mkatychev]

That's a good question, I feel like we would need to know more about tree sitter internals in distinguishing MISSING vs. ERROR I would encourage you to post this here to get a response from the tree-sitter maintainers or I can do so myself come tomorrow in my timezone

[yoshinari-nomura]

Thanks.

I wish I could, but I feel I don't have enough knowledge to ask questions about or understand details.

I can do so myself come tomorrow in my timezone

if you don't mind. thanks.