Fix/series by tag inconsistency with carbon

.github/workflows/release.yml

@@ -32,7 +32,7 @@ jobs:
     - name: Set up Ruby
       uses: ruby/setup-ruby@v1
       with:
-        ruby-version: '2.7' # Version range or exact version of a Ruby version to use, using semvers version range syntax.
+        ruby-version: '3.3' # Version range or exact version of a Ruby version to use, using semvers version range syntax.
     - name: Install packaging dependencies
       run: |
         gem install fpm package_cloud

.github/workflows/tests.yml

@@ -93,14 +93,14 @@ jobs:
     - name: Set up Ruby
       uses: ruby/setup-ruby@v1
       with:
-        ruby-version: '2.7' # Version range or exact version of a Ruby version to use, using semvers version range syntax.
+        ruby-version: '3.3' # Version range or exact version of a Ruby version to use, using semvers version range syntax.
 
+    # gem install dotenv -v 2.8.1 # workaroaund for ruby version 2.7.8.225
     - name: Install packaging dependencies
       run: |
-        gem install dotenv -v 2.8.1 # workaroaund for ruby version 2.7.8.225
         gem install fpm package_cloud
         go install github.com/mitchellh/gox@latest
-
+        
     - name: Check packaging
       run: |
         make DEVEL=1 gox-build fpm-deb fpm-rpm

autocomplete/autocomplete.go

@@ -93,7 +93,7 @@ func (h *Handler) requestExpr(r *http.Request) (*where.Where, *where.Where, map[
 		return wr, pw, usedTags, err
 	}
 
-	wr, pw, err = finder.TaggedWhere(terms)
+	wr, pw, err = finder.TaggedWhere(terms, h.config.FeatureFlags.UseCarbonBehavior, h.config.FeatureFlags.DontMatchMissingTags)
 	if err != nil {
 		return wr, pw, usedTags, err
 	}

config/config.go

@@ -118,6 +118,12 @@ type Common struct {
 	FindCache cache.BytesCache `toml:"-" json:"-"`
 }
 
+// FeatureFlags contains feature flags that significantly change how gch responds to some requests
+type FeatureFlags struct {
+	UseCarbonBehavior    bool `toml:"use-carbon-behaviour" json:"use-carbon-behaviour" comment:"if true, prefers carbon's behaviour on how tags are treated"`
+	DontMatchMissingTags bool `toml:"dont-match-missing-tags" json:"dont-match-missing-tags" comment:"if true, seriesByTag terms containing '!=' or '!=~' operators will not match metrics that don't have the tag at all"`
+}
+
 // IndexReverseRule contains rules to use direct or reversed request to index table
 type IndexReverseRule struct {
 	Suffix   string         `toml:"suffix,omitempty" json:"suffix"  comment:"rule is used when the target suffix is matched"`
@@ -341,15 +347,16 @@ type Debug struct {
 
 // Config is the daemon configuration
 type Config struct {
-	Common     Common             `toml:"common"     json:"common"`
-	Metrics    metrics.Config     `toml:"metrics"    json:"metrics"`
-	ClickHouse ClickHouse         `toml:"clickhouse" json:"clickhouse"`
-	DataTable  []DataTable        `toml:"data-table" json:"data-table" comment:"data tables, see doc/config.md for additional info"`
-	Tags       Tags               `toml:"tags"       json:"tags"       comment:"is not recommended to use, https://github.com/lomik/graphite-clickhouse/wiki/TagsRU" commented:"true"`
-	Carbonlink Carbonlink         `toml:"carbonlink" json:"carbonlink"`
-	Prometheus Prometheus         `toml:"prometheus" json:"prometheus"`
-	Debug      Debug              `toml:"debug"      json:"debug"      comment:"see doc/debugging.md"`
-	Logging    []zapwriter.Config `toml:"logging"    json:"logging"`
+	Common       Common             `toml:"common"        json:"common"`
+	FeatureFlags FeatureFlags       `toml:"feature-flags" json:"feature-flags"`
+	Metrics      metrics.Config     `toml:"metrics"       json:"metrics"`
+	ClickHouse   ClickHouse         `toml:"clickhouse"    json:"clickhouse"`
+	DataTable    []DataTable        `toml:"data-table"    json:"data-table" comment:"data tables, see doc/config.md for additional info"`
+	Tags         Tags               `toml:"tags"          json:"tags"       comment:"is not recommended to use, https://github.com/lomik/graphite-clickhouse/wiki/TagsRU" commented:"true"`
+	Carbonlink   Carbonlink         `toml:"carbonlink"    json:"carbonlink"`
+	Prometheus   Prometheus         `toml:"prometheus"    json:"prometheus"`
+	Debug        Debug              `toml:"debug"         json:"debug"      comment:"see doc/debugging.md"`
+	Logging      []zapwriter.Config `toml:"logging"       json:"logging"`
 }
 
 // New returns *Config with default values

deploy/doc/config.md

@@ -29,6 +29,36 @@ shortTimeoutSec = 300
 findTimeoutSec = 600
 ```
 
+## Feature flags `[feature-flags]`
+
+`use-carbon-behaviour=true`.
+
+- Tagged terms with `=` operator and empty value (e.g. `t=`) match all metrics that don't have that tag.
+
+`dont-match-missing-tags=true`.
+
+- Tagged terms with `!=`, `!=~` operators only match metrics that have that tag.
+
+### Examples
+
+Given tagged metrics:
+```
+metric.two;env=prod
+metric.one;env=stage;dc=mydc1
+metric.one;env=prod;dc=otherdc1
+```
+| Target                      | use-carbon-behaviour | Matched metrics                                   |
+|-----------------------------|----------------------|---------------------------------------------------|
+| seriesByTag('dc=')          | false                | -                                                 |
+| seriesByTag('dc=')          | true                 | metric.two;env=prod                               |
+
+| Target                   | dont-match-missing-tags | Matched metrics                                        |
+|--------------------------|-------------------------|--------------------------------------------------------|
+| seriesByTag('dc!=mydc1') | false                   | metric.two;env=prod<br>metric.one;env=prod;dc=otherdc1 |
+| seriesByTag('dc!=mydc1') | true                    | metric.one;env=prod;dc=otherdc1                        |
+| seriesByTag('dc!=~otherdc') | false                | metric.two;env=prod<br>metric.one;env=stage;dc=mydc1 |
+| seriesByTag('dc!=~otherdc') | true                 | metric.one;env=stage;dc=mydc1                     |
+
 ## ClickHouse `[clickhouse]`
 
 ### URL `url`

doc/config.md

@@ -32,6 +32,36 @@ shortTimeoutSec = 300
 findTimeoutSec = 600
 ```
 
+## Feature flags `[feature-flags]`
+
+`use-carbon-behaviour=true`.
+
+- Tagged terms with `=` operator and empty value (e.g. `t=`) match all metrics that don't have that tag.
+
+`dont-match-missing-tags=true`.
+
+- Tagged terms with `!=`, `!=~` operators only match metrics that have that tag.
+
+### Examples
+
+Given tagged metrics:
+```
+metric.two;env=prod
+metric.one;env=stage;dc=mydc1
+metric.one;env=prod;dc=otherdc1
+```
+| Target                      | use-carbon-behaviour | Matched metrics                                   |
+|-----------------------------|----------------------|---------------------------------------------------|
+| seriesByTag('dc=')          | false                | -                                                 |
+| seriesByTag('dc=')          | true                 | metric.two;env=prod                               |
+
+| Target                   | dont-match-missing-tags | Matched metrics                                        |
+|--------------------------|-------------------------|--------------------------------------------------------|
+| seriesByTag('dc!=mydc1') | false                   | metric.two;env=prod<br>metric.one;env=prod;dc=otherdc1 |
+| seriesByTag('dc!=mydc1') | true                    | metric.one;env=prod;dc=otherdc1                        |
+| seriesByTag('dc!=~otherdc') | false                | metric.two;env=prod<br>metric.one;env=stage;dc=mydc1 |
+| seriesByTag('dc!=~otherdc') | true                 | metric.one;env=stage;dc=mydc1                     |
+
 ## ClickHouse `[clickhouse]`
 
 ### URL `url`
@@ -228,6 +258,12 @@ Only one tag used as filter for index field Tag1, see graphite_tagged table [str
   # offset beetween now and until for select short cache timeout
   short-offset = 0
 
+[feature-flags]
+ # if true, prefers carbon's behaviour on how tags are treated
+ use-carbon-behaviour = false
+ # if true, seriesByTag terms containing '!=' or '!=~' operators will not match metrics that don't have the tag at all
+ dont-match-missing-tags = false
+
 [metrics]
  # graphite relay address
  metric-endpoint = ""

finder/finder.go

@@ -38,7 +38,16 @@ func newPlainFinder(ctx context.Context, config *config.Config, query string, fr
 	var f Finder
 
 	if config.ClickHouse.TaggedTable != "" && strings.HasPrefix(strings.TrimSpace(query), "seriesByTag") {
-		f = NewTagged(config.ClickHouse.URL, config.ClickHouse.TaggedTable, config.ClickHouse.TaggedUseDaily, false, opts, config.ClickHouse.TaggedCosts)
+		f = NewTagged(
+			config.ClickHouse.URL,
+			config.ClickHouse.TaggedTable,
+			config.ClickHouse.TaggedUseDaily,
+			config.FeatureFlags.UseCarbonBehavior,
+			config.FeatureFlags.DontMatchMissingTags,
+			false,
+			opts,
+			config.ClickHouse.TaggedCosts,
+		)
 
 		if len(config.Common.Blacklist) > 0 {
 			f = WrapBlacklist(f, config.Common.Blacklist)
@@ -121,7 +130,16 @@ func FindTagged(ctx context.Context, config *config.Config, terms []TaggedTerm,
 		return Result(plain), nil
 	}
 
-	fnd := NewTagged(config.ClickHouse.URL, config.ClickHouse.TaggedTable, config.ClickHouse.TaggedUseDaily, true, opts, config.ClickHouse.TaggedCosts)
+	fnd := NewTagged(
+		config.ClickHouse.URL,
+		config.ClickHouse.TaggedTable,
+		config.ClickHouse.TaggedUseDaily,
+		config.FeatureFlags.UseCarbonBehavior,
+		config.FeatureFlags.DontMatchMissingTags,
+		true,
+		opts,
+		config.ClickHouse.TaggedCosts,
+	)
 
 	err := fnd.ExecutePrepared(ctx, terms, from, until, stat)
 	if err != nil {

finder/index.go

@@ -178,7 +178,11 @@ func (idx *IndexFinder) Abs(v []byte) []byte {
 }
 
 func (idx *IndexFinder) bodySplit() {
-	idx.rows = bytes.Split(idx.body, []byte{'\n'})
+	if len(idx.body) == 0 {
+		return
+	}
+
+	idx.rows = bytes.Split(bytes.TrimSuffix(idx.body, []byte{'\n'}), []byte{'\n'})
 
 	if idx.useReverse("") {
 		// rotate names for reduce