Filter by ID: a list of ids is not enough to make a query (steps 13 and 14)

[kzangeli]

The problem

Two requests query /entities with id as their only selector:

• 1️⃣3️⃣ Filter Data Entities by ID — id=...&pick=id,type,temperature&format=concise
• 1️⃣4️⃣ Returning data as GeoJSON — id=...&format=concise

Clause 10.4.3.4 of the NGSI-LD API says a query shall carry at least one of

• a. a type selector
• b. a list of attribute names
• c. an NGSI-LD Query
• d. an NGSI-LD GeoQuery
• e. local scope

and raises BadRequestData — too wide query — if none is present. The clause says it in prose too: "it is not possible to retrieve a set of entities by only specifying desired Entity identifiers, without further specifying restrictions on the entities' types or attributes".

pick is not on that list, and is not a selector: it chooses which members of an entity are returned, not which entities are.

The fix

Adds local=true to both, which is clause (e). Two reasons to prefer it over attrs or type:

• it keeps step 13's own point intact — "every id must be unique, so type is not required for this request";
• attrs cannot be combined with pick; they are alternative projection mechanisms and a broker rejects the pair.

A short paragraph explains the rule so the next person editing these requests knows why the parameter is there.

Verified against a running broker: both amended requests return exactly the responses the page documents, including the GeoJSON FeatureCollection for step 14.

The same defect in tutorials.CRUD-Operations step 13 is fixed in a separate PR. Independent of #2 — different lines, no conflict.

🤖 Generated with Claude Code