Add support to return multiple indexes when multiple matches are found #222
Conversation
|
On the surface I like it and I can see various uses for this feature. I'm not sure about the name either, it's kinda long and I don't think the term Hashtag is used anywhere in the docs. But that's a minor thing. Perhaps just One issue might be how it works with modifiers (or mulitpaths). For example, with this document: {
"objectArray":[
{"first": "Dale", "age": 44},
{"first": "Roger", "age": 68},
]
}The path gjson.Result{
Type: 5,
Raw: "[\"Dale\",\"Roger\"]",
Str: "",
Num: 0,
Index: 0,
HashtagIndexes: []int{ 33, 66 },
}While the path gjson.Result{
Type: 5,
Raw: "[\"Roger\",\"Dale\"]",
Str: "",
Num: 0,
Index: 0,
HashtagIndexes: []int{ 11, 41 },
}Because the modifier changed the order of the array, the 11 and 41 no longer applies to the original document. We would either need to remap the indexes to the original document, or maybe we just return nil for the HashtagIndexes array. |
Set to nil when modifiers used
|
Thank you for the feedback! I like the idea of just using the name I've updated the pull request to use |
|
@tidwall Is there anything else you would like me to do to help get this merged? Thanks! |
|
LGTM. I just pushed a new version that includes your PR. :) |
|
Awesome! Thanks :D |
This commit allows for updating values for more "complex" paths like: friends.#(last="Murphy")#.last This is allowed because GJSON now tracks the origin positions of all results (tidwall/gjson#222). This new ability is limited to updating values only. Setting new values that previously did not exist, or deleting values will return an error.
This pull request adds a new field to the
Resultstructure that will contain multiple indexes when multiple matches are found. Currently only a single index is provided that is set to 0 when you provide a query to gather multiple matches (a query to access multiple child paths"friends.#.first"or to find all matches#(...)#). I called the new fieldHashtagIndexesbecause I believe this should only apply when a query uses a hashtag (but definitely open to any other name suggestions).I am working on a project that would benefit from knowing all the indexes, I added a unit test called
TestHashtagIndexesMatchesRawthat is very similar to what I would like to do with this feature. Basically run two separate queries and walk over the results of one and use the indexes to find matches. Thank you for looking over this pull request! Hopefully this is the right approach to achieve this.