Cloudant/Mango selector for deeply nested JSONs
With Cloudant Query (Mango) selector statements, you still need to define an appropriate index before querying. With that in mind, here's your answer:
json-type CQ index
{ "index": { "fields": [ "what_i_want.is_down_here.0" ] }, "type": "json"}Selector against json-type index
{ "selector": { "what_i_want.is_down_here.0": { "i_look_for": "this_tag" }, "what_i_want.is_down_here.0.tag_properties": { "$exists": true } }, "fields": [ "_id", "what_i_want.is_down_here.0.tag_properties" ]}The solution above assumes that you always know/can guarantee the fields you want are within the 0th element of the is_down_here array.
There is another way to answer this question with a different CQ index type. This article explains the differences, and has helpful examples that show querying arrays. Now that you know a little more about the different index types, here's how you'd answer your question with a Lucene search/"text"-type CQ index:
text-type CQ index
{ "index": { "fields": [ {"name": "what_i_want.is_down_here.[]", "type": "string"} ] }, "type": "text"}Selector against text-type index
{ "selector": { "what_i_want.is_down_here": { "$and": [ {"$elemMatch": {"i_look_for": "this_tag"}}, {"$elemMatch": {"tag_properties": {"$exists": true}}} ] } }, "fields": [ "_id", "what_i_want.is_down_here" ]}Read the article and you'll learn that each approach has its tradeoffs: json-type indexes are smaller and less flexible (can only index specific elements); text-type is larger but more flexible (can index all array elements). And from this example, you can also see that the projected values also come with some tradeoffs (projecting specific values vs. the entire array).
More examples in these threads:
If I'm understanding your question properly, there are two supported ways of doing this according to the docs:
{ "what_i_want": { "i_look_for": "this_tag" }}should be equivalent to the abbreviated form:
{ "what_i_want.i_look_for": "this_tag"}