Comment Parsing error

[adamgordonbell]

I think this may be a bug. The following works in a docker file, but not in an Earthfile:

FROM gafiatulin/alpine-sbt

# install docker in docker
RUN set -eux; \
	apk add --no-cache \
		btrfs-progs \
		e2fsprogs \
		e2fsprogs-extra \
		iptables \
		openssl \
		shadow-uidmap \
		# xfsprogs \
		xz \
		pigz \
	; 

Output

line 13:2 token recognition error at: 'xz '
line 13:5 token recognition error at: '\'
line 14:2 token recognition error at: 'pigz '
line 14:7 token recognition error at: '\'
line 15:1 token recognition error at: ';'
line 13:6 no viable alternative at input '\n'
line 14:8 no viable alternative at input '\n'
line 15:2 no viable alternative at input ' \\n\n'
line 13:2 token recognition error at: 'xz '
line 13:5 token recognition error at: '\'
line 14:2 token recognition error at: 'pigz '
line 14:7 token recognition error at: '\'
line 15:1 token recognition error at: ';'
line 13:6 no viable alternative at input '\n'
line 14:8 no viable alternative at input '\n'
line 15:2 no viable alternative at input ' \\n\n'
Error: syntax error: line 13:6 no viable alternative at input '\n'
syntax error: line 14:8 no viable alternative at input '\n'
syntax error: line 15:2 no viable alternative at input ' \\n\n'

[alexcb]

That's interesting that docker supports that, it's definitely related to the inline comment.

I ran it with an older version of earth (before #288 was merged in) and it parsed it fine; however this is what was run:

+base | *cached* --> RUN set -eux; apk add --no-cache btrfs-progs e2fsprogs e2fsprogs-extra iptables openssl shadow-uidmap # xfsprogs xz pigz ;

which had the effect of not installing xz or pigz. Would this be the desired effect or a bug?

I took a quick look at python for inspiration, and noticed that both following code examples are invalid:

x = "a" \
    #hello \
    "b"
print(x)

and

x = "a" \
    \ #hello
    "b"
print(x)

[vladaionescu]

FTR, we talked about this internally and we think the best approach is to treat comments as if they're not there. Also, it might be a good idea if blank lines are also ignored - meaning that you can have blank lines between continuations.

So all of these should work then:

Blank line would be ignored, so continuation still works.

... \

... \
...

Similar to above, but not as obvious, comment is treated like a blank line, and blank lines are ignored, so continuation still works.

... \
    #hello \
...\
...

Comment after continuation character is ok:

... \
    \ #hello
... \
...

[rafaeelaudibert]

Hello guys! Maybe I could tackle this issue?

Already tracked it down to stmtWords on EarthParser.g4 not allowing empty lines and/or comment lines in between statements.

I believe we could change this rule to use a new type of optional blank lines which would go towards the direction proposed by @vladaionescu

[vladaionescu]

SGTM - it may require testing, experimentation and fiddling. I'll assign this to you.

[adamgordonbell]

This problem also prevents commenting on long RUN statements which are required by WITH DOCKER, making it hard to document the build process.

WITH DOCKER 
        RUN docker-compose up -d & \                                                                                            # start docker compose
            MIX_ENV=test mix deps.compile && \                                                                                  # in parellel start compiling
            while ! sqlcmd -S tcp:localhost,1433 -U sa -P 'some!Password' -Q "SELECT 1" > /dev/null 2>&1; do sleep 1; done; \   # is sql server up?
            while ! mysqladmin ping --host=localhost --port=3306 --protocol=TCP --silent; do sleep 1; done; \                   # is mysql up? 
            while ! pg_isready --host=localhost --port=5432 --quiet; do sleep 1; done; \                                        # is pg up? 
            mix test --include database                                                                                         # lets test 
    END