Add nilsimsa2 order mode (will replace nilsimsa after more testing)

include/dwarfs/inode_manager.h

@@ -25,14 +25,14 @@
 #include <functional>
 #include <memory>
 
-#include "dwarfs/options.h"
-
 namespace dwarfs {
 
 class inode;
 class logger;
 class script;
 
+struct file_order_options;
+
 class inode_manager {
  public:
   using inode_cb = std::function<void(std::shared_ptr<inode> const&)>;
@@ -43,8 +43,9 @@ class inode_manager {
 
   size_t count() const { return impl_->count(); }
 
-  void order_inodes(std::shared_ptr<script> scr, file_order_mode file_order,
-                    uint32_t first_inode, inode_cb const& fn) {
+  void order_inodes(std::shared_ptr<script> scr,
+                    file_order_options const& file_order, uint32_t first_inode,
+                    inode_cb const& fn) {
     impl_->order_inodes(std::move(scr), file_order, first_inode, fn);
   }
 
@@ -56,9 +57,9 @@ class inode_manager {
 
     virtual std::shared_ptr<inode> create_inode() = 0;
     virtual size_t count() const = 0;
-    virtual void
-    order_inodes(std::shared_ptr<script> scr, file_order_mode file_order,
-                 uint32_t first_inode, inode_cb const& fn) = 0;
+    virtual void order_inodes(std::shared_ptr<script> scr,
+                              file_order_options const& file_order,
+                              uint32_t first_inode, inode_cb const& fn) = 0;
     virtual void for_each_inode(
         std::function<void(std::shared_ptr<inode> const&)> const& fn) const = 0;
   };

include/dwarfs/options.h

@@ -54,10 +54,23 @@ struct inode_options {
   bool needs_scan() const { return with_similarity || with_nilsimsa; }
 };
 
-enum class file_order_mode { NONE, PATH, SCRIPT, SIMILARITY, NILSIMSA };
+enum class file_order_mode {
+  NONE,
+  PATH,
+  SCRIPT,
+  SIMILARITY,
+  NILSIMSA,
+  NILSIMSA2
+};
+
+struct file_order_options {
+  file_order_mode mode{file_order_mode::NONE};
+  int nilsimsa_depth{10000};
+  int nilsimsa_limit{250};
+};
 
 struct scanner_options {
-  file_order_mode file_order{file_order_mode::NONE};
+  file_order_options file_order;
   std::optional<uint16_t> uid;
   std::optional<uint16_t> gid;
   std::optional<uint64_t> timestamp;

src/dwarfs/inode_manager.cpp

@@ -35,6 +35,7 @@
 #include "dwarfs/logger.h"
 #include "dwarfs/mmif.h"
 #include "dwarfs/nilsimsa.h"
+#include "dwarfs/options.h"
 #include "dwarfs/os_access.h"
 #include "dwarfs/script.h"
 #include "dwarfs/similarity.h"
@@ -159,8 +160,8 @@ class inode_manager_ : public inode_manager::impl {
 
   size_t count() const override { return inodes_.size(); }
 
-  void order_inodes(std::shared_ptr<script> scr, file_order_mode file_order,
-                    uint32_t first_inode,
+  void order_inodes(std::shared_ptr<script> scr,
+                    file_order_options const& file_order, uint32_t first_inode,
                     inode_manager::inode_cb const& fn) override;
 
   void
@@ -213,6 +214,10 @@ class inode_manager_ : public inode_manager::impl {
   void order_inodes_by_nilsimsa(inode_manager::inode_cb const& fn,
                                 uint32_t inode_no);
 
+  void order_inodes_by_nilsimsa2(inode_manager::inode_cb const& fn,
+                                 uint32_t inode_no,
+                                 file_order_options const& file_order);
+
   void number_inodes(size_t first_no) {
     for (auto& i : inodes_) {
       i->set_num(first_no++);
@@ -225,9 +230,9 @@ class inode_manager_ : public inode_manager::impl {
 
 template <typename LoggerPolicy>
 void inode_manager_<LoggerPolicy>::order_inodes(
-    std::shared_ptr<script> scr, file_order_mode file_order,
+    std::shared_ptr<script> scr, file_order_options const& file_order,
     uint32_t first_inode, inode_manager::inode_cb const& fn) {
-  switch (file_order) {
+  switch (file_order.mode) {
   case file_order_mode::NONE:
     log_.info() << "keeping inode order";
     break;
@@ -265,15 +270,22 @@ void inode_manager_<LoggerPolicy>::order_inodes(
     auto ti = log_.timed_info();
     order_inodes_by_nilsimsa(fn, first_inode);
     ti << count() << " inodes ordered";
-    break;
+    return;
+  }
+
+  case file_order_mode::NILSIMSA2: {
+    log_.info() << "ordering " << count()
+                << " inodes using nilsimsa2 similarity...";
+    auto ti = log_.timed_info();
+    order_inodes_by_nilsimsa2(fn, first_inode, file_order);
+    ti << count() << " inodes ordered";
+    return;
   }
   }
 
-  if (file_order != file_order_mode::NILSIMSA) {
-    log_.info() << "assigning file inodes...";
-    number_inodes(first_inode);
-    for_each_inode(fn);
-  }
+  log_.info() << "assigning file inodes...";
+  number_inodes(first_inode);
+  for_each_inode(fn);
 }
 
 template <typename LoggerPolicy>
@@ -399,6 +411,85 @@ void inode_manager_<LoggerPolicy>::order_inodes_by_nilsimsa(
   }
 }
 
+template <typename LoggerPolicy>
+void inode_manager_<LoggerPolicy>::order_inodes_by_nilsimsa2(
+    inode_manager::inode_cb const& fn, uint32_t inode_no,
+    file_order_options const& file_order) {
+  auto count = inodes_.size();
+
+  std::vector<std::shared_ptr<inode>> inodes;
+  inodes.swap(inodes_);
+  inodes_.reserve(count);
+  std::vector<uint32_t> index;
+  index.resize(count);
+  std::iota(index.begin(), index.end(), 0);
+
+  auto finalize_inode = [&]() {
+    inodes_.push_back(std::move(inodes[index.back()]));
+    index.pop_back();
+    inodes_.back()->set_num(inode_no++);
+    fn(inodes_.back());
+  };
+
+  auto empty = std::partition(index.begin(), index.end(),
+                              [&](auto i) { return inodes[i]->size() > 0; });
+
+  if (empty != index.end()) {
+    assert(empty + 1 == index.end());
+    finalize_inode();
+  }
+
+  if (!index.empty()) {
+    const int depth = file_order.nilsimsa_depth;
+    const int limit = file_order.nilsimsa_limit;
+
+    log_.info() << "nilsimsa: depth=" << depth << ", limit=" << limit;
+
+    std::sort(index.begin(), index.end(), [&](auto a, auto b) {
+      auto const& ia = *inodes[a];
+      auto const& ib = *inodes[b];
+      return (ia.size() < ib.size() ||
+              (ia.size() == ib.size() && ia.any()->path() < ib.any()->path()));
+    });
+
+    finalize_inode();
+
+    while (!index.empty()) {
+      auto* ref_hash = inodes_.back()->nilsimsa_similarity_hash().data();
+
+      int max_sim = 0;
+      int max_sim_ix = 0;
+
+      int end = int(index.size()) > depth ? index.size() - depth : 0;
+
+      for (int i = index.size() - 1; i >= end; --i) {
+        auto sim = dwarfs::nilsimsa_similarity(
+            ref_hash, inodes[index[i]]->nilsimsa_similarity_hash().data());
+
+        if (sim > max_sim) {
+          max_sim = sim;
+          max_sim_ix = i;
+
+          if (max_sim >= limit) {
+            break;
+          }
+        }
+      }
+
+      log_.trace() << max_sim << " @ " << max_sim_ix << "/" << index.size();
+
+      std::rotate(index.begin() + max_sim_ix, index.begin() + max_sim_ix + 1,
+                  index.end());
+
+      finalize_inode();
+    }
+  }
+
+  if (count != inodes_.size()) {
+    throw std::runtime_error("internal error: nilsimsa ordering failed");
+  }
+}
+
 inode_manager::inode_manager(logger& lgr)
     : impl_(make_unique_logging_object<impl, inode_manager_, logger_policies>(
           lgr)) {}

src/mkdwarfs.cpp

@@ -40,7 +40,6 @@
 #include <unistd.h>
 
 #include <boost/algorithm/string.hpp>
-#include <boost/any.hpp>
 #include <boost/program_options.hpp>
 
 #include <folly/Conv.h>
@@ -96,7 +95,8 @@ const std::map<std::string, file_order_mode> order_choices{
     {"script", file_order_mode::SCRIPT},
 #endif
     {"similarity", file_order_mode::SIMILARITY},
-    {"nilsimsa", file_order_mode::NILSIMSA}};
+    {"nilsimsa", file_order_mode::NILSIMSA},
+    {"nilsimsa2", file_order_mode::NILSIMSA2}};
 
 const std::map<std::string, uint32_t> time_resolutions{
     {"sec", 1},
@@ -109,20 +109,6 @@ const std::map<std::string, uint32_t> time_resolutions{
 
 namespace dwarfs {
 
-void validate(boost::any& v, const std::vector<std::string>& values,
-              file_order_mode*, int) {
-  using namespace boost::program_options;
-
-  validators::check_first_occurrence(v);
-
-  auto it = order_choices.find(validators::get_single_string(values));
-  if (it == order_choices.end()) {
-    throw validation_error(validation_error::invalid_option_value);
-  }
-
-  v = boost::any(it->second);
-}
-
 class script_options : public options_interface {
  public:
   script_options(logger& lgr, po::variables_map& vm, scanner_options& opts,
@@ -133,7 +119,7 @@ class script_options : public options_interface {
       , force_similarity_(force_similarity) {}
 
   void set_order(file_order_mode order_mode, set_mode mode = DEFAULT) override {
-    set(opts_.file_order, order_mode, "order", mode);
+    set(opts_.file_order.mode, order_mode, "order", mode);
   }
 
   void
@@ -289,7 +275,7 @@ int mkdwarfs(int argc, char** argv) {
   block_manager::config cfg;
   std::string path, output, window_sizes, memory_limit, script_arg, compression,
       schema_compression, metadata_compression, log_level, timestamp,
-      time_resolution;
+      time_resolution, order;
   size_t num_workers, max_scanner_workers;
   bool recompress = false, no_progress = false;
   unsigned level;
@@ -356,8 +342,8 @@ int mkdwarfs(int argc, char** argv) {
         po::value<std::string>(&time_resolution)->default_value("sec"),
         resolution_desc.c_str())
     ("order",
-        po::value<file_order_mode>(&options.file_order)
-            ->default_value(file_order_mode::SIMILARITY, "similarity"),
+        po::value<std::string>(&order)
+            ->default_value("similarity"),
         order_desc.c_str())
 #ifdef DWARFS_HAVE_PYTHON
     ("script",
@@ -490,6 +476,40 @@ int mkdwarfs(int argc, char** argv) {
     window_sizes = defaults.window_sizes;
   }
 
+  std::vector<std::string> order_opts;
+  boost::split(order_opts, order, boost::is_any_of(":"));
+  if (auto it = order_choices.find(order_opts.front());
+      it != order_choices.end()) {
+    options.file_order.mode = it->second;
+    if (order_opts.size() > 1) {
+      if (options.file_order.mode != file_order_mode::NILSIMSA2) {
+        throw std::runtime_error(
+            fmt::format("file order mode '{}' does not support options",
+                        order_opts.front()));
+      }
+      if (order_opts.size() > 3) {
+        throw std::runtime_error(fmt::format(
+            "too many options for file order mode '{}'", order_opts.front()));
+      }
+      options.file_order.nilsimsa_limit = folly::to<int>(order_opts[1]);
+      if (options.file_order.nilsimsa_limit < 0 ||
+          options.file_order.nilsimsa_limit > 255) {
+        throw std::runtime_error(
+            fmt::format("limit ({}) out of range for '{}' (0..255)",
+                        options.file_order.nilsimsa_limit, order_opts.front()));
+      }
+      if (order_opts.size() > 2) {
+        options.file_order.nilsimsa_depth = folly::to<int>(order_opts[2]);
+        if (options.file_order.nilsimsa_depth < 0) {
+          throw std::runtime_error(fmt::format(
+              "depth ({}) cannot be negative for '{}'", order_opts.front()));
+        }
+      }
+    }
+  } else {
+    throw std::runtime_error("invalid file order mode: " + order);
+  }
+
   size_t mem_limit = parse_size_with_unit(memory_limit);
 
   std::vector<std::string> wsv;
@@ -544,7 +564,7 @@ int mkdwarfs(int argc, char** argv) {
     script->configure(script_opts);
   }
 
-  if (options.file_order == file_order_mode::SCRIPT && !script) {
+  if (options.file_order.mode == file_order_mode::SCRIPT && !script) {
     throw std::runtime_error(
         "--order=script can only be used with a valid --script option");
   }
@@ -591,9 +611,11 @@ int mkdwarfs(int argc, char** argv) {
     ti << "filesystem rewritten";
   } else {
     options.inode.with_similarity =
-        force_similarity || options.file_order == file_order_mode::SIMILARITY;
+        force_similarity ||
+        options.file_order.mode == file_order_mode::SIMILARITY;
     options.inode.with_nilsimsa =
-        options.file_order == file_order_mode::NILSIMSA;
+        options.file_order.mode == file_order_mode::NILSIMSA ||
+        options.file_order.mode == file_order_mode::NILSIMSA2;
 
     scanner s(lgr, wg_scanner, cfg, entry_factory::create(),
               std::make_shared<os_access_posix>(), std::move(script), options);

test/dwarfs.cpp

@@ -199,7 +199,7 @@ void basic_end_to_end_test(std::string const& compressor,
   cfg.blockhash_window_size.push_back(1 << 10);
   cfg.block_size_bits = block_size_bits;
 
-  options.file_order = file_order;
+  options.file_order.mode = file_order;
   options.with_devices = with_devices;
   options.with_specials = with_specials;
   options.inode.with_similarity = file_order == file_order_mode::SIMILARITY;