Wrong json umarshal bug for model value

https://github.com/prometheus/common/blob/49b36038aead362ce78f09337dc5872000bd048a/model/value.go#L51

I found a strange bug, the response data unmarshalled by prom client-go is wrong by chance.

I use the queryRange API provided by prometheus client go as followings

and i print the time series data returned by the queryRange api, it has unexpected data value. it is very large and unexpected.

the value logged function is below, i use json marshal directly to marshal the returned data by api, i am sure that this is the data returned by prometheus client-go api queryRange function:

and then i use curl curl -G --data-urlencode 'query=irate(container_cpu_usage_seconds_total{container!="POD",namespace="ns-prjchlvc-1544041-production",pod=~"^monitor-server-siliconvalley-150001.*$",container="second-monitor-server"}[3m])' --data-urlencode 'start=1658205420.000' --data-urlencode 'end=1658865360.000' --data-urlencode 'step=60' http://127.0.0.1:9090/api/v1/query_range to query the prometheus directly, it returned normal data as expected. you can see the following data timestamp and value, it is what we want. but the above image is use the client-go api, with the same param, it has unexpected value at the same timestamp.

So i think the json unmarshal has some bug the prometheus client go impelmented

Start at model/value.go around line 51 and inspect how query-range response values are unmarshalled. Reproduce the report with the provided query, timestamps, and step, then compare the client result with the direct curl response. Done means the unmarshalled timestamp and value data match the Prometheus response.