Add timestamp_extractor to simplify timestamp extraction

This is a simpler program which has fewer moving parts.
Sometimes, --skip_order_validation with log_cat isn't aggressive enough
to make the log readable because it is that busted.  timestamp_extractor
is good enough to parse those logs.

Change-Id: I8e5250cd149ef2f1b86384d61064d49da41f42c2
diff --git a/aos/events/logging/BUILD b/aos/events/logging/BUILD
index 827aa99..b88478d 100644
--- a/aos/events/logging/BUILD
+++ b/aos/events/logging/BUILD
@@ -208,7 +208,6 @@
         "//aos:configuration",
         "//aos:init",
         "//aos:json_to_flatbuffer",
-        "//aos/events:shm_event_loop",
         "//aos/events:simulated_event_loop",
         "@com_github_gflags_gflags//:gflags",
         "@com_github_google_glog//:glog",
@@ -216,6 +215,24 @@
 )
 
 cc_binary(
+    name = "timestamp_extractor",
+    srcs = [
+        "timestamp_extractor.cc",
+    ],
+    target_compatible_with = ["@platforms//os:linux"],
+    visibility = ["//visibility:public"],
+    deps = [
+        ":logfile_utils",
+        "//aos:configuration",
+        "//aos:init",
+        "//aos/events:simulated_event_loop",
+        "//aos/network:multinode_timestamp_filter",
+        "@com_github_gflags_gflags//:gflags",
+        "@com_github_google_glog//:glog",
+    ],
+)
+
+cc_binary(
     name = "log_edit",
     srcs = [
         "log_edit.cc",
diff --git a/aos/events/logging/logger.cc b/aos/events/logging/logger.cc
index a8975d2..493f09a 100644
--- a/aos/events/logging/logger.cc
+++ b/aos/events/logging/logger.cc
@@ -1111,7 +1111,7 @@
   remapped_configuration_ = event_loop_factory_->configuration();
   filters_ =
       std::make_unique<message_bridge::MultiNodeNoncausalOffsetEstimator>(
-          event_loop_factory_, logged_configuration(),
+          event_loop_factory_->configuration(), logged_configuration(),
           FLAGS_skip_order_validation,
           chrono::duration_cast<chrono::nanoseconds>(
               chrono::duration<double>(FLAGS_time_estimation_buffer_seconds)));
@@ -1264,15 +1264,6 @@
 
   if (FLAGS_timestamps_to_csv) {
     filters_->Start(event_loop_factory);
-    std::fstream s("/tmp/timestamp_noncausal_starttime.csv", s.trunc | s.out);
-    CHECK(s.is_open());
-    for (std::unique_ptr<State> &state : states_) {
-      s << state->event_loop()->node()->name()->string_view() << ", "
-        << std::setprecision(12) << std::fixed
-        << chrono::duration<double>(state->monotonic_now().time_since_epoch())
-               .count()
-        << "\n";
-    }
   }
 }
 
diff --git a/aos/events/logging/timestamp_extractor.cc b/aos/events/logging/timestamp_extractor.cc
new file mode 100644
index 0000000..5ec7025
--- /dev/null
+++ b/aos/events/logging/timestamp_extractor.cc
@@ -0,0 +1,177 @@
+#include <iostream>
+#include <string>
+#include <vector>
+
+#include "aos/events/logging/logfile_sorting.h"
+#include "aos/events/logging/logfile_utils.h"
+#include "aos/init.h"
+#include "aos/network/multinode_timestamp_filter.h"
+#include "gflags/gflags.h"
+
+DECLARE_bool(timestamps_to_csv);
+DEFINE_bool(skip_order_validation, false,
+            "If true, ignore any out of orderness in replay");
+
+namespace aos::logger {
+
+namespace chrono = std::chrono;
+
+std::string LogFileVectorToString(std::vector<logger::LogFile> log_files) {
+  std::stringstream ss;
+  for (const auto f : log_files) {
+    ss << f << "\n";
+  }
+  return ss.str();
+}
+
+int Main(int argc, char **argv) {
+  const std::vector<std::string> unsorted_logfiles = FindLogs(argc, argv);
+  const std::vector<LogFile> log_files = SortParts(unsorted_logfiles);
+
+  CHECK_GT(log_files.size(), 0u);
+  // Validate that we have the same config everwhere.  This will be true if
+  // all the parts were sorted together and the configs match.
+  const Configuration *config = nullptr;
+  for (const LogFile &log_file : log_files) {
+    if (config == nullptr) {
+      config = log_file.config.get();
+    } else {
+      CHECK_EQ(config, log_file.config.get());
+    }
+  }
+
+  CHECK(configuration::MultiNode(config))
+      << ": Timestamps only make sense in a multi-node world.";
+
+  // Now, build up all the TimestampMapper classes to read and sort the data.
+  std::vector<std::unique_ptr<TimestampMapper>> mappers;
+
+  for (const Node *node : configuration::GetNodes(config)) {
+    std::vector<LogParts> filtered_parts =
+        FilterPartsForNode(log_files, node->name()->string_view());
+
+    // Confirm that all the parts are from the same boot if there are enough
+    // parts to not be from the same boot.
+    if (!filtered_parts.empty()) {
+      for (size_t i = 1; i < filtered_parts.size(); ++i) {
+        CHECK_EQ(filtered_parts[i].source_boot_uuid,
+                 filtered_parts[0].source_boot_uuid)
+            << ": Found parts from different boots "
+            << LogFileVectorToString(log_files);
+      }
+
+      // Filter the parts relevant to each node when building the mapper.
+      mappers.emplace_back(
+          std::make_unique<TimestampMapper>(std::move(filtered_parts)));
+    } else {
+      mappers.emplace_back(nullptr);
+    }
+  }
+
+  // Now, build up the estimator used to solve for time.
+  message_bridge::MultiNodeNoncausalOffsetEstimator multinode_estimator(
+      config, config, FLAGS_skip_order_validation, chrono::seconds(0));
+
+  // To make things more like the logger and faster, cache the node + channel ->
+  // filter mapping in a set of vectors.
+  std::vector<std::vector<message_bridge::NoncausalOffsetEstimator *>> filters;
+  filters.resize(configuration::NodesCount(config));
+
+  for (const Node *node : configuration::GetNodes(config)) {
+    const size_t node_index = configuration::GetNodeIndex(config, node);
+    filters[node_index].resize(config->channels()->size(), nullptr);
+    for (size_t channel_index = 0; channel_index < config->channels()->size();
+         ++channel_index) {
+      const Channel *channel = config->channels()->Get(channel_index);
+
+      if (!configuration::ChannelIsSendableOnNode(channel, node) &&
+          configuration::ChannelIsReadableOnNode(channel, node)) {
+        // We've got a message which is being forwarded to this node.
+        const Node *source_node = configuration::GetNode(
+            config, channel->source_node()->string_view());
+        filters[node_index][channel_index] =
+            multinode_estimator.GetFilter(node, source_node);
+      }
+    }
+  }
+
+  // Now, read all the timestamps for each node.  This is simpler than the
+  // logger on purpose.  It loads in *all* the timestamps in 1 go per node,
+  // ignoring memory usage.
+  for (const Node *node : configuration::GetNodes(config)) {
+    LOG(INFO) << "Reading all data for " << node->name()->string_view();
+    const size_t node_index = configuration::GetNodeIndex(config, node);
+    TimestampMapper *timestamp_mapper = mappers[node_index].get();
+    if (timestamp_mapper == nullptr) {
+      continue;
+    }
+    while (true) {
+      TimestampedMessage *m = timestamp_mapper->Front();
+      if (m == nullptr) {
+        break;
+      }
+
+      if (m->monotonic_remote_time != monotonic_clock::min_time) {
+        // Got a forwarding timestamp!
+        message_bridge::NoncausalOffsetEstimator *filter =
+            filters[node_index][m->channel_index];
+        CHECK(filter != nullptr);
+
+        filter->Sample(node, m->monotonic_event_time, m->monotonic_remote_time);
+
+        // Call the correct method depending on if we are the forward or
+        // reverse direction here.
+        if (m->monotonic_timestamp_time != monotonic_clock::min_time) {
+          // TODO(austin): This assumes that this timestamp is only logged on
+          // the node which sent the data.  That is correct for now, but should
+          // be explicitly checked somewhere.
+          filter->ReverseSample(node, m->monotonic_event_time,
+                                m->monotonic_timestamp_time);
+        }
+      }
+      timestamp_mapper->PopFront();
+    }
+  }
+
+  // Don't get clever. Use the first time as the start time.  Note: this is
+  // different than how log_cat and others work.
+  std::optional<std::tuple<distributed_clock::time_point,
+                           std::vector<monotonic_clock::time_point>>>
+      next_timestamp = multinode_estimator.NextTimestamp();
+  CHECK(next_timestamp);
+  LOG(INFO) << "Starting at:";
+  for (const Node *node : configuration::GetNodes(config)) {
+    const size_t node_index = configuration::GetNodeIndex(config, node);
+    LOG(INFO) << "  " << node->name()->string_view() << " -> "
+              << std::get<1>(*next_timestamp)[node_index];
+  }
+
+  multinode_estimator.Start(std::get<1>(*next_timestamp));
+
+  // As we pull off all the timestamps, the time problem is continually solved,
+  // filling in the CSV files.
+  while (true) {
+    std::optional<std::tuple<distributed_clock::time_point,
+                             std::vector<monotonic_clock::time_point>>>
+        next_timestamp = multinode_estimator.NextTimestamp();
+    if (!next_timestamp) {
+      break;
+    }
+  }
+
+  return 0;
+}
+
+}  // namespace aos::logger
+
+int main(int argc, char **argv) {
+  FLAGS_timestamps_to_csv = true;
+  gflags::SetUsageMessage(
+      "Usage:\n"
+      "  timestamp_extractor [args] logfile1 logfile2 ...\n\nThis program "
+      "dumps out all the timestamps from a set of log files for plotting.  Use "
+      "--skip_order_validation to skip any time estimation problems we find.");
+  aos::InitGoogle(&argc, &argv);
+
+  return aos::logger::Main(argc, argv);
+}
diff --git a/aos/events/logging/timestamp_plot.gnuplot b/aos/events/logging/timestamp_plot.gnuplot
index 566448a..7ad62da 100755
--- a/aos/events/logging/timestamp_plot.gnuplot
+++ b/aos/events/logging/timestamp_plot.gnuplot
@@ -28,6 +28,6 @@
      samplefile21 using 1:(-$2) title 'sample 2-1', \
      noncausalfile12 using 1:3 title 'nc 1-2' with lines, \
      noncausalfile21 using 1:(-$3) title 'nc 2-1' with lines, \
-     offsetfile using ((column(node1_index) - node1_start_time + (column(node2_index) - node2_start_time)) / 2):(column(node2_index) - column(node1_index)) title 'filter 2-1' with lines
+     offsetfile using ((column(node1_index) - node1_start_time + (column(node2_index) - node2_start_time)) / 2):(column(node2_index) - column(node1_index)) title 'filter 2-1' with linespoints
 
 pause -1