-
Notifications
You must be signed in to change notification settings - Fork 3.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Literal Expressions in LogQL #1677
Conversation
Codecov Report
@@ Coverage Diff @@
## master #1677 +/- ##
==========================================
+ Coverage 62.81% 63.43% +0.61%
==========================================
Files 112 121 +9
Lines 8741 9011 +270
==========================================
+ Hits 5491 5716 +225
- Misses 2837 2877 +40
- Partials 413 418 +5
|
7d9174d
to
e33ec36
Compare
Don't merge this until I add support for a response result type for
and I should add another |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Great work !
Hey @cyriltovena I updated this to properly use The ugly:
I'll integration test this tomorrow or Monday. [edit] I found a way to improve the evaluator code. This hinges on proving that a binary expression necessarily produces a vector evaluator. This is assured because all binary ops where both legs are literals can be recursively reduced at construction Once that's assured, we simply check if the root node of an AST is a literal and handle that edge case. Otherwise, all subsequent evaluators will produce vectors which allows us to reduce the type complexity therein. |
@@ -288,6 +288,21 @@ func TestEngine_NewInstantQuery(t *testing.T) { | |||
promql.Sample{Point: promql.Point{T: 60 * 1000, V: 0.2}, Metric: labels.Labels{labels.Label{Name: "app", Value: "fuzz"}}}, | |||
}, | |||
}, | |||
{ | |||
// healthcheck |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we have a test for Literal + Vector ?
{`topk(3,count_over_time({foo="bar"}[5m])) by (foo,bar)`, []int{TOPK, OPEN_PARENTHESIS, IDENTIFIER, COMMA, COUNT_OVER_TIME, OPEN_PARENTHESIS, OPEN_BRACE, IDENTIFIER, EQ, STRING, CLOSE_BRACE, DURATION, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS, BY, OPEN_PARENTHESIS, IDENTIFIER, COMMA, IDENTIFIER, CLOSE_PARENTHESIS}}, | ||
{`bottomk(10,sum(count_over_time({foo="bar"}[5m])) by (foo,bar))`, []int{BOTTOMK, OPEN_PARENTHESIS, IDENTIFIER, COMMA, SUM, OPEN_PARENTHESIS, COUNT_OVER_TIME, OPEN_PARENTHESIS, OPEN_BRACE, IDENTIFIER, EQ, STRING, CLOSE_BRACE, DURATION, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS, BY, OPEN_PARENTHESIS, IDENTIFIER, COMMA, IDENTIFIER, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS}}, | ||
{`topk(3,count_over_time({foo="bar"}[5m])) by (foo,bar)`, []int{TOPK, OPEN_PARENTHESIS, NUMBER, COMMA, COUNT_OVER_TIME, OPEN_PARENTHESIS, OPEN_BRACE, IDENTIFIER, EQ, STRING, CLOSE_BRACE, DURATION, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS, BY, OPEN_PARENTHESIS, IDENTIFIER, COMMA, IDENTIFIER, CLOSE_PARENTHESIS}}, | ||
{`bottomk(10,sum(count_over_time({foo="bar"}[5m])) by (foo,bar))`, []int{BOTTOMK, OPEN_PARENTHESIS, NUMBER, COMMA, SUM, OPEN_PARENTHESIS, COUNT_OVER_TIME, OPEN_PARENTHESIS, OPEN_BRACE, IDENTIFIER, EQ, STRING, CLOSE_BRACE, DURATION, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS, BY, OPEN_PARENTHESIS, IDENTIFIER, COMMA, IDENTIFIER, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS}}, | ||
{`sum(max(rate({foo="bar"}[5m])) by (foo,bar)) by (foo)`, []int{SUM, OPEN_PARENTHESIS, MAX, OPEN_PARENTHESIS, RATE, OPEN_PARENTHESIS, OPEN_BRACE, IDENTIFIER, EQ, STRING, CLOSE_BRACE, DURATION, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS, BY, OPEN_PARENTHESIS, IDENTIFIER, COMMA, IDENTIFIER, CLOSE_PARENTHESIS, CLOSE_PARENTHESIS, BY, OPEN_PARENTHESIS, IDENTIFIER, CLOSE_PARENTHESIS}}, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Would be nice to have some new tests here too.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
What
Introduces numeric literals!
Description
Binary arithmetic operators are now defined between two literals (scalars), a literal and a vector, and two vectors.
Between two literals, the behavior is obvious: they evaluate to another literal that is the result of the operator applied to both scalar operands (1 + 1 = 2).
Between a vector and a literal, the operator is applied to the value of every data sample in the vector. E.g. if a time series vector is multiplied by 2, the result is another vector in which every sample value of the original vector is multiplied by 2.
Between two vectors, a binary arithmetic operator is applied to each entry in the left-hand side vector and its matching element in the right-hand vector. The result is propagated into the result vector with the grouping labels becoming the output label set. Entries for which no matching entry in the right-hand vector can be found are not part of the result.
Examples
Implement a health check with a simple query:
1 + 1
Double the rate of a a log stream's entries:
sum(rate({app="foo"})) * 2