More tests for backslash behaviour #438
Conversation
| @@ -1,2 +1,3 @@ | |||
| node1 | |||
| \ | |||
There was a problem hiding this comment.
Based on the filename I figured this test was missing something...
There was a problem hiding this comment.
This is allowed per spec, but I'm surprised by that.
line-space doesn't contain escline, only node-space has escline
All whitespace between nodes is line-space, so on first sight this test would be invalid, but…
base-node starts with slashdash? type? node-space* string which means this is valid
node1
\
node2
but this isn't
(lorem)node1
\
(lorem)node2
I personally always read the start of base-node as slashdash? (type node-space*)? string, as I interpreted the node-space to be meant as space between the type and name. Apparently that's also how I implemented this, as my parser fails this test atm.
There was a problem hiding this comment.
...I noticed this recently in kdl-rs too and it made me wonder whether escline should be part of line-space.
There was a problem hiding this comment.
what in the spec allows this, btw?
There was a problem hiding this comment.
Ah I see:
(type) \
node
is allowed and (type) is optional, therefore
\
node
is allowed.
Quite separately,
/- \
(type)node
and even
/-
\
(type)node
are allowed because /- can always be followed by node-space, even when not in a node.
IMHO, this is too confusing and escline belongs in line-space.
|
Oh, I almost sent a similar PR :) |
| @@ -0,0 +1 @@ | |||
| \ | |||
There was a problem hiding this comment.
Now I'm confused: is this actually allowed? I think it should be (just allow all whitespace to be escaped even if it's meaning less) but right now I'm not sure what the spec actually says right now
|
I think a test using non-literal whitespace for prefix could also be useful, e.g. +++ b/tests/test_cases/input/multiline_string_nonliteral_prefix_fail.kdl
@@ -0,0 +1,5 @@
+node """
+ hey
+ \severyone
+ how goes?
+ """ |
* Release KDL 2.0.0 * fix grammar for multiline quoted strings to allow escaped whitespace on closing line * Add unicode-space to raw string * Remove nonexistent equals-sign from the grammar (#435) * fix multiline string tests * grammar: fix disallowed-keyword-identifiers and string-character (#436) * Back out "fix multiline string tests" This backs out commit 0c5604b. * add extra javascript implementation (#437) * reword interaction multiline + whitespace escape (#439) * More tests for backslash behaviour (#438) * More tests for baskslash behaviour * Incorrect example of escaped final newline * Test with non-literal indent * Make line-space a superset of node-space (#440) * Allow escline everywhere * escline tests * Always escape \ inside single quotes in the grammar text (#441) to match the other uses of it and the metalanguage description below * Add tests for mandatory whitespace between arguments or properties (#442) * Add an optional version marker (#444) * Add version marker to the grammer * Add version marker to the Changelog * Update SPEC.md Co-authored-by: eilvelia <[email protected]> * add a mandatory newline after the version marker * add mandatory space between version number --------- Co-authored-by: eilvelia <[email protected]> * Fix a changelog line erroneously truncated in #444 (#445) * fix: move vertical tab to the line-breaking whitespace to match Unicode (#446) * add vertical tab change test * final tweaks before release --------- Co-authored-by: eilvelia <[email protected]> Co-authored-by: Bram Gotink <[email protected]> Co-authored-by: Thomas Jollans <[email protected]> Co-authored-by: Evgeny <[email protected]>
Tests for the cases brought up by @eilvelia and @bgotink in #434, as well some more tests for (non-string) escline