Fix all XPathMatcherTest TODO's + handle nested elements with the same name #4532
Conversation
| @@ -92,18 +92,8 @@ public boolean matches(Cursor cursor) { | |||
| if (index < 0) { | |||
| return false; | |||
| } | |||
| if (part.startsWith("@")) { // is attribute selector | |||
| partWithCondition = part; | |||
| tagForCondition = i > 0 ? path.get(i - 1) : path.get(i); | |||
There was a problem hiding this comment.
Using i straight up in path.get() was completely wrong since path is reverted while parts is not – pathIndex should always be used.
Note that if a part starts with @, it MUST be the last part – I wanted to check it here, but then I realized all ifs actually result in the same code (and conditions matching sub-elements were not handled…).
| } | ||
| } | ||
| partWithCondition = part; | ||
| tagForCondition = path.get(pathIndex); | ||
| } else if (i < path.size() && i > 0 && parts[i - 1].endsWith("]")) { |
There was a problem hiding this comment.
I don’t understand what’s the purpose of this if. All tests still pass if I remove it but I’m not sure whether I’m missing an edge case or it’s really unneeded.
- paths with multiple occurrences of // - // preceding @Attribute
| } else { | ||
| Collections.reverse(path); | ||
|
|
||
| // Deal with the two forward slashes in the expression; works, but I'm not proud of it. | ||
| if (expression.contains("//") && !expression.contains("://") && Arrays.stream(parts).anyMatch(StringUtils::isBlank)) { |
There was a problem hiding this comment.
Removing :// does not seem to break anything, but fixes XPaths with both // and a namespace URL condition.
I don’t think the first condition is still relevant either, however the isBlank check is inconsistent with the indexOf("") on the next line. I think it should use isEmpty() like I did to compute tagMatchingParts.
Note that the first condition is insufficient on its own because of URLs again – or actually any value that could contain double slashes in a condition.
What's changed?
//What's your motivation?
Needed to use such XPaths.
Anything in particular you'd like reviewers to focus on?
I will put comments on the bits of code I cleaned up.
There are still limitations with XPath that do not start with
/or that start with//, in particular, they don’t support//in the middle and there were not tests for that.Have you considered any alternatives or workarounds?
I think the implementation would really still benefit from a lot of refactoring, or even a full rewrite:
/is completely dissociated from those starting with//or no/, which causes different limitations for the two.I think it would make sense to represent the XPath as a structured object instead of a
String[]…Any additional context
As a user, it’s not really clear what is supported and what is not without looking at the unit test…
Checklist