Make the histograms more flexible.

[nageru] / metrics.cpp

diff --git a/metrics.cpp b/metrics.cpp
index 3b3d1eb2217e6b2aadbdb4e922811de864161d12..3468ed766b17cbd11b987fcaac03ddfd8414dd51 100644 (file)
--- a/metrics.cpp
+++ b/metrics.cpp
@@ -2,6 +2,7 @@
 
 #include <assert.h>
 
+#include <algorithm>
 #include <locale>
 #include <sstream>
 
@@ -57,17 +58,18 @@ void Metrics::add(const string &name, const vector<pair<string, string>> &labels
        types[name] = type;
 }
 
-void Metrics::add_histogram(const string &name, const vector<pair<string, string>> &labels, atomic<int64_t> *first_bucket_location, atomic<double> *sum_location, size_t num_elements)
+void Metrics::add(const string &name, const vector<pair<string, string>> &labels, Histogram *location)
 {
-       Histogram histogram;
-       histogram.name = name;
-       histogram.labels = labels;
-       histogram.first_bucket_location = first_bucket_location;
-       histogram.sum_location = sum_location;
-       histogram.num_elements = num_elements;
+       Metric metric;
+       metric.data_type = DATA_TYPE_HISTOGRAM;
+       metric.name = name;
+       metric.labels = labels;
+       metric.location_histogram = location;
 
        lock_guard<mutex> lock(mu);
-       histograms.push_back(histogram);
+       metrics.push_back(metric);
+       assert(types.count(name) == 0 || types[name] == TYPE_HISTOGRAM);
+       types[name] = TYPE_HISTOGRAM;
 }
 
 string Metrics::serialize() const
@@ -80,6 +82,8 @@ string Metrics::serialize() const
        for (const auto &name_and_type : types) {
                if (name_and_type.second == TYPE_GAUGE) {
                        ss << "# TYPE nageru_" << name_and_type.first << " gauge\n";
+               } else if (name_and_type.second == TYPE_HISTOGRAM) {
+                       ss << "# TYPE nageru_" << name_and_type.first << " histogram\n";
                }
        }
        for (const Metric &metric : metrics) {
@@ -87,28 +91,72 @@ string Metrics::serialize() const
 
                if (metric.data_type == DATA_TYPE_INT64) {
                        ss << name << " " << metric.location_int64->load() << "\n";
-               } else {
+               } else if (metric.data_type == DATA_TYPE_DOUBLE) {
                        ss << name << " " << metric.location_double->load() << "\n";
+               } else {
+                       ss << metric.location_histogram->serialize(metric.name, metric.labels);
                }
        }
-       for (const Histogram &histogram : histograms) {
-               ss << "# TYPE nageru_" << histogram.name << " histogram\n";
-
-               int64_t count = 0;
-               for (size_t i = 0; i < histogram.num_elements; ++i) {
-                       char buf[16];
-                       snprintf(buf, sizeof(buf), "%lu", i);
-                       vector<pair<string, string>> labels = histogram.labels;
-                       labels.emplace_back("le", buf);
-
-                       int64_t val = histogram.first_bucket_location[i].load();
-                       count += val;
-                       ss << serialize_name(histogram.name + "_bucket", labels) << " " << count << "\n";
-               }
 
-               ss << serialize_name(histogram.name + "_sum", histogram.labels) << " " << histogram.sum_location->load() << "\n";
-               ss << serialize_name(histogram.name + "_count", histogram.labels) << " " << count << "\n";
+       return ss.str();
+}
+
+void Histogram::init(const vector<double> &bucket_vals)
+{
+       this->num_buckets = bucket_vals.size();
+       buckets.reset(new Bucket[num_buckets]);
+       for (size_t i = 0; i < num_buckets; ++i) {
+               buckets[i].val = bucket_vals[i];
+       }
+}
+
+void Histogram::init_uniform(size_t num_buckets)
+{
+       this->num_buckets = num_buckets;
+       buckets.reset(new Bucket[num_buckets]);
+       for (size_t i = 0; i < num_buckets; ++i) {
+               buckets[i].val = i;
        }
+}
+
+void Histogram::count_event(double val)
+{
+       Bucket ref_bucket;
+       ref_bucket.val = val;
+       auto it = lower_bound(buckets.get(), buckets.get() + num_buckets, ref_bucket,
+               [](const Bucket &a, const Bucket &b) { return a.val < b.val; });
+       if (it == buckets.get() + num_buckets) {
+               ++count_after_last_bucket;
+       } else {
+               ++it->count;
+       }
+       // Non-atomic add, but that's fine, since there are no concurrent writers.
+       sum = sum + val;
+}
+
+string Histogram::serialize(const string &name, const vector<pair<string, string>> &labels) const
+{
+       stringstream ss;
+       ss.imbue(locale("C"));
+       ss.precision(20);
+
+       int64_t count = 0;
+       for (size_t bucket_idx = 0; bucket_idx < num_buckets; ++bucket_idx) {
+               stringstream le_ss;
+               le_ss.imbue(locale("C"));
+               le_ss.precision(20);
+               le_ss << buckets[bucket_idx].val;
+               vector<pair<string, string>> bucket_labels = labels;
+               bucket_labels.emplace_back("le", le_ss.str());
+
+               count += buckets[bucket_idx].count.load();
+               ss << serialize_name(name + "_bucket", bucket_labels) << " " << count << "\n";
+       }
+
+       count += count_after_last_bucket.load();
+
+       ss << serialize_name(name + "_sum", labels) << " " << sum.load() << "\n";
+       ss << serialize_name(name + "_count", labels) << " " << count << "\n";
 
        return ss.str();
 }