Remove trailing spaces

CMakeLists.txt

@@ -25,7 +25,7 @@ list(APPEND CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake/modules)
 
 # We need boost
 find_package(Boost 1.36.0 REQUIRED COMPONENTS
-  program_options 
+  program_options
   system
   thread
   unit_test_framework
@@ -36,7 +36,7 @@ find_package(Boost 1.36.0 REQUIRED COMPONENTS
 
 # Define where include files live
 include_directories(
-  ${PROJECT_SOURCE_DIR} 
+  ${PROJECT_SOURCE_DIR}
   ${Boost_INCLUDE_DIRS}
 )
 

lm/CMakeLists.txt

@@ -12,7 +12,7 @@ add_definitions(-DKENLM_MAX_ORDER=${KENLM_MAX_ORDER})
 #    that should be included in the kenlm library,
 #        (this excludes any unit test files)
 #    you should add them to the following list:
-set(KENLM_LM_SOURCE 
+set(KENLM_LM_SOURCE
 	bhiksha.cc
 	binary_format.cc
 	config.cc
@@ -31,7 +31,7 @@ set(KENLM_LM_SOURCE
 )
 
 
-# Group these objects together for later use. 
+# Group these objects together for later use.
 #
 # Given add_library(foo OBJECT ${my_foo_sources}),
 # refer to these objects as $<TARGET_OBJECTS:foo>

lm/builder/CMakeLists.txt

@@ -11,7 +11,7 @@
 #    in case this variable is referenced by CMake files in the parent directory,
 #    we prefix all files with ${CMAKE_CURRENT_SOURCE_DIR}.
 #
-set(KENLM_BUILDER_SOURCE 
+set(KENLM_BUILDER_SOURCE
 		${CMAKE_CURRENT_SOURCE_DIR}/adjust_counts.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/corpus_count.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/initial_probabilities.cc
@@ -21,7 +21,7 @@ set(KENLM_BUILDER_SOURCE
 	)
 
 
-# Group these objects together for later use. 
+# Group these objects together for later use.
 #
 # Given add_library(foo OBJECT ${my_foo_sources}),
 # refer to these objects as $<TARGET_OBJECTS:foo>

lm/common/model_buffer.cc

@@ -17,7 +17,7 @@ const char kMetadataHeader[] = "KenLM intermediate binary file";
 ModelBuffer::ModelBuffer(StringPiece file_base, bool keep_buffer, bool output_q)
   : file_base_(file_base.data(), file_base.size()), keep_buffer_(keep_buffer), output_q_(output_q),
     vocab_file_(keep_buffer ? util::CreateOrThrow((file_base_ + ".vocab").c_str()) : util::MakeTemp(file_base_)) {}
-  
+
 ModelBuffer::ModelBuffer(StringPiece file_base)
   : file_base_(file_base.data(), file_base.size()), keep_buffer_(false) {
   const std::string full_name = file_base_ + ".kenlm_intermediate";

lm/filter/CMakeLists.txt

@@ -11,14 +11,14 @@
 #    in case this variable is referenced by CMake files in the parent directory,
 #    we prefix all files with ${CMAKE_CURRENT_SOURCE_DIR}.
 #
-set(KENLM_FILTER_SOURCE 
+set(KENLM_FILTER_SOURCE
 		${CMAKE_CURRENT_SOURCE_DIR}/arpa_io.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/phrase.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/vocab.cc
 	)
 
 
-# Group these objects together for later use. 
+# Group these objects together for later use.
 #
 # Given add_library(foo OBJECT ${my_foo_sources}),
 # refer to these objects as $<TARGET_OBJECTS:foo>

lm/filter/arpa_io.cc

@@ -45,7 +45,7 @@ bool IsEntirelyWhiteSpace(const StringPiece &line) {
   return true;
 }
 
-ARPAOutput::ARPAOutput(const char *name, size_t buffer_size) 
+ARPAOutput::ARPAOutput(const char *name, size_t buffer_size)
   : file_backing_(util::CreateOrThrow(name)), file_(file_backing_.get(), buffer_size) {}
 
 void ARPAOutput::ReserveForCounts(std::streampos reserve) {

lm/interpolate/tune_derivatives.cc

@@ -88,7 +88,7 @@ Accum Derivatives(Instances &in, const Vector &weights, Vector &gradient, Matrix
         - interp_uni(word) * Z_epsilon * weighted_backoffs * ln_p_i_backed * ln_p_i_backed.transpose();
     }
 
-    Accum Z_context = 
+    Accum Z_context =
       weighted_backoffs * Z_epsilon * (1.0 - sum_x_p_I) // Back off and unnormalize the unigrams for which there is no extension.
       + unnormalized_sum_x_p_I_full; // Add the extensions.
     sum_ln_Z_context += log(Z_context);

lm/interpolate/tune_derivatives_test.cc

@@ -47,7 +47,7 @@ namespace {
 
 BOOST_AUTO_TEST_CASE(Small) {
   MockInstances mock;
-  
+
   {
     // Three vocabulary words plus <s>, two models.
     Matrix unigrams(4, 2);

lm/interpolate/tune_instances.cc

@@ -2,7 +2,7 @@
  * instance is an n-gram in the tuning file.  To tune towards these, we want
  * the correct probability p_i(w_n | w_1^{n-1}) from each model as well as
  * all the denominators p_i(v | w_1^{n-1}) that appear in normalization.
- * 
+ *
  * In other words, we filter the models to only those n-grams whose context
  * appears in the tuning data.  This can be divided into two categories:
  * - All unigrams.  This goes into Instances::ln_unigrams_
@@ -11,7 +11,7 @@
  *   w_1^{n-1}v since that is what will be used for the probability.
  * Because there is a large number of extensions (we tried keeping them in RAM
  * and ran out), the streaming framework is used to keep track of extensions
- * and sort them so they can be streamed in.  Downstream code 
+ * and sort them so they can be streamed in.  Downstream code
  * (tune_derivatives.hh) takes a stream of extensions ordered by tuning
  * instance, the word v, and the model the extension came from.
  */
@@ -186,7 +186,7 @@ class JointOrderCallback {
         std::size_t full_order_minus_1,
         ContextMap &contexts,
         util::stream::Stream &out,
-        const InstanceMatch *base_instance) 
+        const InstanceMatch *base_instance)
       : full_order_minus_1_(full_order_minus_1),
         contexts_(contexts),
         out_(out),
@@ -231,13 +231,13 @@ class JointOrderCallback {
     // Mapping is constant but values are being manipulated to tell them about
     // n-grams.
     ContextMap &contexts_;
-    
+
     // Reused variable.  model is set correctly.
     InitialExtension ext_;
 
     util::stream::Stream &out_;
-    
-    const InstanceMatch *const base_instance_;    
+
+    const InstanceMatch *const base_instance_;
 };
 
 // This populates the ln_unigrams_ matrix.  It can (and should for efficiency)
@@ -247,7 +247,7 @@ class ReadUnigrams {
     explicit ReadUnigrams(Matrix::ColXpr out) : out_(out) {}
 
     // Read renumbered unigrams, fill with <unk> otherwise.
-    void Run(const util::stream::ChainPosition &position) { 
+    void Run(const util::stream::ChainPosition &position) {
       NGramStream<ProbBackoff> stream(position);
       assert(stream);
       Accum unk = stream->Value().prob * M_LN10;
@@ -315,7 +315,7 @@ class IdentifyTuning : public EnumerateVocab {
 // Store information about the first iteration.
 class ExtensionsFirstIteration {
   public:
-    explicit ExtensionsFirstIteration(std::size_t instances, std::size_t models, std::size_t max_order, util::stream::Chain &extension_input, const util::stream::SortConfig &config) 
+    explicit ExtensionsFirstIteration(std::size_t instances, std::size_t models, std::size_t max_order, util::stream::Chain &extension_input, const util::stream::SortConfig &config)
       : backoffs_by_instance_(new std::vector<Matrix>(instances)), sort_(extension_input, config) {
       // Initialize all the backoff matrices to zeros.
       for (std::vector<Matrix>::iterator i = backoffs_by_instance_->begin(); i != backoffs_by_instance_->end(); ++i) {

lm/interpolate/tune_instances.hh

@@ -79,7 +79,7 @@ class Instances {
 
     // backoffs_(instance, model) is the backoff all the way to unigrams.
     BackoffMatrix ln_backoffs_;
-    
+
     // neg_correct_sum_(model) = -\sum_{instances} ln p_{model}(correct(instance) | context(instance)).
     // This appears as a term in the gradient.
     Vector neg_ln_correct_sum_;

lm/interpolate/tune_instances_test.cc

@@ -47,7 +47,7 @@ BOOST_AUTO_TEST_CASE(Toy) {
 
   BOOST_CHECK_EQUAL(1, inst.BOS());
   const Matrix &ln_unigrams = inst.LNUnigrams();
-  
+
   // <unk>=0
   BOOST_CHECK_CLOSE(-0.90309 * M_LN10, ln_unigrams(0, 0), 0.001);
   BOOST_CHECK_CLOSE(-1 * M_LN10, ln_unigrams(0, 1), 0.001);
@@ -64,15 +64,15 @@ BOOST_AUTO_TEST_CASE(Toy) {
   BOOST_CHECK_CLOSE(-0.90309 * M_LN10, ln_unigrams(4, 0), 0.001); // <unk>
   BOOST_CHECK_CLOSE(-0.7659168 * M_LN10, ln_unigrams(4, 1), 0.001);
   // too lazy to do b = 5.
-  
+
   // Two instances:
   // <s> predicts c
   // <s> c predicts </s>
   BOOST_REQUIRE_EQUAL(2, inst.NumInstances());
   BOOST_CHECK_CLOSE(-0.30103 * M_LN10, inst.LNBackoffs(0)(0), 0.001);
   BOOST_CHECK_CLOSE(-0.30103 * M_LN10, inst.LNBackoffs(0)(1), 0.001);
 
-  
+
   // Backoffs of <s> c
   BOOST_CHECK_CLOSE(0.0, inst.LNBackoffs(1)(0), 0.001);
   BOOST_CHECK_CLOSE((-0.30103 - 0.30103) * M_LN10, inst.LNBackoffs(1)(1), 0.001);
@@ -88,7 +88,7 @@ BOOST_AUTO_TEST_CASE(Toy) {
   // <s> b from model 0
   // c </s> from model 1
   // Magic probabilities come from querying the models directly.
-  
+
   // <s> a from model 0
   BOOST_REQUIRE(stream);
   BOOST_CHECK_EQUAL(0, stream->instance);

lm/interpolate/tune_matrix.hh

@@ -8,7 +8,7 @@
 #pragma GCC diagnostic pop
 
 namespace lm { namespace interpolate {
- 
+
 typedef Eigen::MatrixXf Matrix;
 typedef Eigen::VectorXf Vector;
 

util/CMakeLists.txt

@@ -11,30 +11,30 @@
 # Because we do not set PARENT_SCOPE in the following definition,
 #    CMake files in the parent directory won't be able to access this variable.
 #
-set(KENLM_UTIL_SOURCE 
-		bit_packing.cc 
-		ersatz_progress.cc 
-		exception.cc 
-		file.cc 
-		file_piece.cc 
+set(KENLM_UTIL_SOURCE
+		bit_packing.cc
+		ersatz_progress.cc
+		exception.cc
+		file.cc
+		file_piece.cc
 		float_to_string.cc
 		integer_to_string.cc
-		mmap.cc 
-		murmur_hash.cc 
+		mmap.cc
+		murmur_hash.cc
 		parallel_read.cc
-		pool.cc 
-		read_compressed.cc 
-		scoped.cc 
+		pool.cc
+		read_compressed.cc
+		scoped.cc
     spaces.cc
-		string_piece.cc 
+		string_piece.cc
 		usage.cc
 	)
 
 # This directory has children that need to be processed
 add_subdirectory(double-conversion)
 add_subdirectory(stream)
 
-# Group these objects together for later use. 
+# Group these objects together for later use.
 add_library(kenlm_util ${KENLM_UTIL_DOUBLECONVERSION_SOURCE} ${KENLM_UTIL_STREAM_SOURCE} ${KENLM_UTIL_SOURCE})
 
 AddExes(EXES probing_hash_table_benchmark

util/double-conversion/CMakeLists.txt

@@ -14,7 +14,7 @@
 #    when this variable is referenced by CMake files in the parent directory,
 #    we prefix all files with ${CMAKE_CURRENT_SOURCE_DIR}.
 #
-set(KENLM_UTIL_DOUBLECONVERSION_SOURCE 
+set(KENLM_UTIL_DOUBLECONVERSION_SOURCE
 		${CMAKE_CURRENT_SOURCE_DIR}/bignum-dtoa.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/bignum.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/cached-powers.cc

util/fake_ostream.hh

@@ -14,13 +14,13 @@ namespace util {
 
 /* Like std::ostream but without being incredibly slow.
  * Supports most of the built-in types except for long double.
- * 
+ *
  * The FakeOStream class is intended to be inherited from.  The inherting class
  * should provide:
  * public:
  *   Derived &flush();
  *   Derived &write(const void *data, std::size_t length);
- * 
+ *
  * private: or protected:
  *   friend class FakeOStream;
  *   char *Ensure(std::size_t amount);
@@ -29,7 +29,7 @@ namespace util {
  * The Ensure function makes enough space for an in-place write and returns
  * where to write.  The AdvanceTo function happens after the write, saying how
  * much was actually written.
- * 
+ *
  * Precondition:
  * amount <= kToStringMaxBytes for in-place writes.
  */

util/file.cc

@@ -158,7 +158,7 @@ std::size_t PartialRead(int fd, void *to, std::size_t amount) {
     DWORD ret;
     HANDLE file_handle = reinterpret_cast<HANDLE>(_get_osfhandle(fd));
     DWORD larger_size = static_cast<DWORD>(std::min<std::size_t>(kMaxDWORD, amount));
-    DWORD smaller_size = 28672; // Received reports that 31346 worked but higher values did not. This rounds down to the nearest multiple of 4096, the page size. 
+    DWORD smaller_size = 28672; // Received reports that 31346 worked but higher values did not. This rounds down to the nearest multiple of 4096, the page size.
     if (!ReadFile(file_handle, to, larger_size, &ret, NULL))
     {
         DWORD last_error = GetLastError();

util/pool.hh

@@ -36,7 +36,7 @@ class Pool {
         void *new_base = More(new_total);
         std::memcpy(new_base, base, new_total - additional);
         base = new_base;
-      } 
+      }
     }
 
     void FreeAll();

util/stream/CMakeLists.txt

@@ -14,7 +14,7 @@
 #    when this variable is referenced by CMake files in the parent directory,
 #    we prefix all files with ${CMAKE_CURRENT_SOURCE_DIR}.
 #
-set(KENLM_UTIL_STREAM_SOURCE 
+set(KENLM_UTIL_STREAM_SOURCE
 		${CMAKE_CURRENT_SOURCE_DIR}/chain.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/count_records.cc
 		${CMAKE_CURRENT_SOURCE_DIR}/io.cc