Is there a reason why we are not covering testing matrix/vector for Query API and queryRange API? Is it because that's not possible in the test suite?

I think the reason is that we do not fill the TSDB with any data for these simple tests. Therefore, all the queries are just using PromQL that still returns something with an empty TSDB.

There are ways to create a fully fledged vector with multiple elements. Not sure what's the best, but here is my spontaneous attempt:

label_replace(vector(42), "foo", "bar", "", "") or label_replace(vector(3.1415), "dings", "bums", "", "")

This returns a vector with two elements, so you could try out limit=1 to see if only one is returned. (And same for a range query.)

In different news: I'll be off for my Xmas vacations. Back 2025-01-07. In the meantime, it would be cool if @machine424 and @bboreham could say if this is doing the right thing. (I think it does, modulo the warnings handling, see above. So we only need the tests beyond that.)

There are ways to create a fully fledged vector with multiple elements. Not sure what's the best, but here is my spontaneous attempt:

label_replace(vector(42), "foo", "bar", "", "") or label_replace(vector(3.1415), "dings", "bums", "", "")

This returns a vector with two elements, so you could try out limit=1 to see if only one is returned. (And same for a range query.)

I couldn't find where I could fill the TSDB with testing data. Added some tests for query and query_range checking warningCount.

My point was that with a test query like label_replace(vector(42), "foo", "bar", "", "") or label_replace(vector(3.1415), "dings", "bums", "", "") you can create a vector with multiple elements without the need to fill the test TSDB with data.

The test case for the query:

label_replace(vector(42), "foo", "bar", "", "") or label_replace(vector(3.1415), "dings", "bums", "", "")

Failed due to a mismatch in the timestamps. The test expected a hardcoded timestamp of 0, but during testing we assign the current timestamp (in Unix epoch format) to generated values by default during execution. Prometheus does not allow you to control or ignore timestamps directly in queries.

The test comparison failed because the timestamps in the actual response did not match the expected value.

+               {
+                       endpoint: api.query,
+                       query: url.Values{
+                               "query": []string{
+                                       `label_replace(vector(42), "foo", "bar", "", "") or label_replace(vector(3.1415), "dings", "bums", "", "")`,
+                               },
+                       },
+                       responseAsJSON: `{
+               "resultType": "vector",
+               "result": [
+                       {
+                               "metric": {
+                                       "foo": "bar"
+                               },
+                               "value": [0, "42"]
+                       },
+                       {
+                               "metric": {
+                                       "dings": "bums"
+                               },
+                               "value": [0, "3.1415"]
+                       }
+               ]
+       }`,
+               },

Can we ignore the timestamps in this test case?

❯ go test ./web/api/v1
--- FAIL: TestEndpoints (0.08s)
    --- FAIL: TestEndpoints/local (0.03s)
        --- FAIL: TestEndpoints/local/run_7_query_"query=label_replace%28vector%2842%29%2C+%22foo%22%2C+%22bar%22%2C+%22%22%2C+%22%22%29+or+label_replace%28vector%283.1415%29%2C+%22dings%22%2C+%22bums%22%2C+%22%22%2C+%22%22%29" (0.00s)
            --- FAIL: TestEndpoints/local/run_7_query_"query=label_replace%28vector%2842%29%2C+%22foo%22%2C+%22bar%22%2C+%22%22%2C+%22%22%29+or+label_replace%28vector%283.1415%29%2C+%22dings%22%2C+%22bums%22%2C+%22%22%2C+%22%22%29"/GET (0.00s)
                api_test.go:3676: 
                    	Error Trace:	/home/vandit/Github-Forks/prometheus/web/api/v1/api_test.go:3676
                    	Error:      	Not equal: 
                    	           	expected: map[string]interface {}{"result":[]interface {}{map[string]interface {}{"metric":map[string]interface {}{"foo":"bar"}, "value":[]interface {}{0, "42"}}, map[string]interface {}{"metric":map[string]interface {}{"dings":"bums"}, "value":[]interface {}{0, "3.1415"}}}, "resultType":"vector"}
                    	           	actual  : map[string]interface {}{"result":[]interface {}{map[string]interface {}{"metric":map[string]interface {}{"foo":"bar"}, "value":[]interface {}{1.734882542759e+09, "42"}}, map[string]interface {}{"metric":map[string]interface {}{"dings":"bums"}, "value":[]interface {}{1.734882542759e+09, "3.1415"}}}, "resultType":"vector"}
                    	           	
                    	           	Diff:
                    	           	--- Expected
                    	           	+++ Actual
                    	           	@@ -7,3 +7,3 @@
                    	           	   (string) (len=5) "value": ([]interface {}) (len=2) {
                    	           	-    (float64) 0,
                    	           	+    (float64) 1.734882542759e+09,
                    	           	    (string) (len=2) "42"
                    	           	@@ -16,3 +16,3 @@
                    	           	   (string) (len=5) "value": ([]interface {}) (len=2) {
                    	           	-    (float64) 0,
                    	           	+    (float64) 1.734882542759e+09,
                    	           	    (string) (len=6) "3.1415"
                    	Test:       	TestEndpoints/local/run_7_query_"query=label_replace%28vector%2842%29%2C+%22foo%22%2C+%22bar%22%2C+%22%22%2C+%22%22%29+or+label_replace%28vector%283.1415%29%2C+%22dings%22%2C+%22bums%22%2C+%22%22%2C+%22%22%29"/GET

I've pushed the changes with tests (in 587c690) that simulate a vector using label_replace. The values match as expected, but the timestamps are different causing the tests to fail.
Should we ignore the timestamp differences, or is there a way to make them match in code?

Sorry, tests are still failing. Seems like a simple fix.

Sorry, tests are still failing. Seems like a simple fix.

Fixed.

🎉 thank you very much.