Build: Remove Eigen patch from code base

Pivotal tracker: 102425066

Apache license recommends MPL code to not be updated directly in the
apache software. Hence, we move the Eigen patch to a separate repository.

We also update the UseLATEX.cmake file to eliminate some
cmake warnings regarding missing dependencies.

doc/design/CMakeLists.txt

@@ -32,7 +32,7 @@ if(LATEX_BINARY AND BIBTEX_BINARY AND MAKEINDEX_BINARY)
         BIBFILES ../literature.bib
         INPUTS
             ${DESIGN_DOC_MODULES}
-        MANGLE_TARGET_NAMES
+        EXCLUDE_FROM_ALL
     )
 else()
     message(STATUS "A complete LaTeX installation could not be found. "

src/CMakeLists.txt

@@ -18,6 +18,10 @@ set(BITBUCKET_BASE_URL
     "${MADLIB_REDIRECT_PREFIX}https://bitbucket.org"
     CACHE STRING
     "Base URL for Bitbucket projects. May be overridden for testing purposes.")
+set(GITHUB_MADLIB_BASE_URL
+    "${MADLIB_REDIRECT_PREFIX}https://github.com/madlib"
+    CACHE STRING
+    "Base URL for MADlib Github projects. May be overridden for testing purposes.")
 
 # Boost might not be present on the system (or simply too old). In this case, we
 # download the following version (unless it is already present in
@@ -47,16 +51,16 @@ endif (NOT BOOST_TAR_SOURCE)
 # specify an alternative path to the Eigen tarball:
 # -DEIGEN_TAR_SOURCE=/path/to/eigen-x.x.x.tar.gz
 
-set(EIGEN_VERSION "3.2.2")
-set(EIGEN_TAR_MD5 f8817c258b1495cb474d412e528aedb6)
+set(EIGEN_VERSION "branches/3.2")
+set(EIGEN_URL "${GITHUB_MADLIB_BASE_URL}/eigen/archive/${EIGEN_VERSION}.tar.gz")
+set(EIGEN_TAR_MD5 52cd37725d0a1852325d9325674423fa)
 set(EIGEN_MPL2_ONLY TRUE)
 
-set(EIGEN_TAR "eigen-${EIGEN_VERSION}.tar.gz")
-set(EIGEN_URL "${BITBUCKET_BASE_URL}/eigen/eigen/get/${EIGEN_VERSION}.tar.gz")
-
 if(NOT EIGEN_TAR_SOURCE)
+    STRING(REGEX REPLACE "/" "-" EIGEN_VERSION_NO_SLASH ${EIGEN_VERSION})
+    set(EIGEN_TAR "eigen-${EIGEN_VERSION_NO_SLASH}.tar.gz")
     find_file(EIGEN_TAR_SOURCE ${EIGEN_TAR}
-        PATHS ${MAD_THIRD_PARTY}/downloads)
+              PATHS ${MAD_THIRD_PARTY}/downloads)
 endif(NOT EIGEN_TAR_SOURCE)
 
 if(NOT EIGEN_TAR_SOURCE)
@@ -129,15 +133,10 @@ endif(Boost_FOUND)
 
 
 # -- Third-party dependencies: Download the C++ linear-algebra library Eigen ---
-
-# FIXME: Eigen is a third-party source that is patched in-place. Other
-# third-party headers are patched in the patch directory.
 ExternalProject_Add(EP_eigen
     PREFIX ${MAD_THIRD_PARTY}
     DOWNLOAD_DIR ${MAD_THIRD_PARTY}/downloads
     URL ${EIGEN_TAR_SOURCE}
-    URL_MD5 ${EIGEN_TAR_MD5}
-    PATCH_COMMAND "${CMAKE_CURRENT_SOURCE_DIR}/patch/Eigen.sh"
     CMAKE_COMMAND /usr/bin/env echo Ignored: cmake
     BUILD_COMMAND /usr/bin/env echo Ignored: make
     INSTALL_COMMAND /usr/bin/env echo Ignored: make

src/modules/convex/task/l1.hpp

@@ -1,5 +1,5 @@
 /* ----------------------------------------------------------------------- *//**
- * 
+ *
  * @file l1.hpp
  *
  *//* ----------------------------------------------------------------------- */
@@ -24,19 +24,19 @@ class L1 {
     typedef Model model_type;
 
     static void gradient(
-            const model_type                    &model,
-            const double                        &lambda,
-            model_type                          &gradient);
+            const model_type                &model,
+            const double                    &lambda,
+            model_type                      &gradient);
 
     static void clipping(
-            model_type                          &incrModel,
-            const double                        &lambda, 
-            const int                           &n_tuples,
-            const double                        &stepsize);
+            model_type                      &incrModel,
+            const double                    &lambda,
+            const int                       &n_tuples,
+            const double                    &stepsize);
 
     static double loss(
-            const model_type                    &model, 
-            const double                        &lambda);
+            const model_type                &model,
+            const double                    &lambda);
 };
 
 template <class Model>
@@ -58,7 +58,7 @@ template <class Model>
 void
 L1<Model>::clipping(
         model_type                          &incrModel,
-        const double                        &lambda, 
+        const double                        &lambda,
         const int                           &n_tuples,
         const double                        &stepsize) {
     // implement the Clipping method mentioned in Tsuruoka et al. 2009
@@ -73,13 +73,13 @@ L1<Model>::clipping(
 }
 
 template <class Model>
-double 
+double
 L1<Model>::loss(
-        const model_type                    &model, 
+        const model_type                    &model,
         const double                        &lambda) {
     double s = 0.;
     for (Index i = 0; i < model.size(); i++) {
-        s += abs(model(i));
+        s += std::abs(model(i));
     }
     return lambda * s;
 }

src/modules/lda/lda.cpp

@@ -378,9 +378,9 @@ AnyType lda_count_topic_sfunc::run(AnyType & args)
         // to store a voc_size x (topic_num+1) integer matrix in
         // bigint[] (the +1 is for a flag of ceiling the count),
         // we need padding if the size is odd.
-        // 1. when voc_size * (topic_num + 1) is (2n+1), gives (n+1) 
-        // 2. when voc_size * (topic_num + 1) is (2n), gives (n) 
-        int dims[1] = {(voc_size * (topic_num + 1) + 1) * sizeof(int32_t) / sizeof(int64_t)};
+        // 1. when voc_size * (topic_num + 1) is (2n+1), gives (n+1)
+        // 2. when voc_size * (topic_num + 1) is (2n), gives (n)
+        int dims[1] = {static_cast<int>( (voc_size * (topic_num + 1) + 1) * sizeof(int32_t) / sizeof(int64_t) )};
         int lbs[1] = {1};
         state = madlib_construct_md_array(
             NULL, NULL, 1, dims, lbs, INT8TI.oid, INT8TI.len, INT8TI.byval,
@@ -407,7 +407,6 @@ AnyType lda_count_topic_sfunc::run(AnyType & args)
             word_index++;
         }
     }
-
     return state;
 }
 

src/ports/postgres/modules/linalg/matrix_ops.py_in

@@ -459,7 +459,6 @@ def _validate_block(matrix, matrix_args):
             "Matrix error: Invalid block column - array expected")
     _assert(not table_is_empty(matrix),
             "Matrix error: Input table {0} is empty".format(matrix))
-# ------------------------------------------------------------------------------
 
 # ------------------------------------------------------------------------------
 # -- Transformation operations -------------------------------------------------
@@ -563,7 +562,6 @@ def matrix_densify(schema_madlib, matrix_in, in_args, matrix_out, out_args):
             FROM generate_series(1, {col_dim}) AS val
         ) t2
         """.format(**locals()))
-# ------------------------------------------------------------------------------
 
 # ------------------------------------------------------------------------------
 # -- Element-wise operations ---------------------------------------------------
@@ -822,7 +820,6 @@ def matrix_elem_mult(schema_madlib, matrix_a, a_args,
     matrix_elem_op(schema_madlib, matrix_a, a_args,
                    matrix_b, b_args, matrix_out, out_args,
                    elem_ops)
-# ------------------------------------------------------------------------------
 
 # ------------------------------------------------------------------------------
 # -- Block operations ------------------------------------------------------
@@ -1027,7 +1024,6 @@ def matrix_unblockize(schema_madlib, matrix_in, in_args, matrix_out, out_args):
         m4_ifdef(`__POSTGRESQL__', `',
             `DISTRIBUTED BY ({out_args[row]})')
         """.format(**locals()))
-# ------------------------------------------------------------------------------
 
 # ------------------------------------------------------------------------------
 # -- Visitor operations --------------------------------------------------------
@@ -1145,7 +1141,6 @@ def _matrix_extract_col_dense(schema_madlib, matrix_in, in_args, index=1):
     _assert(len(r) == 1,
             "Matrix error: Invalid row index ({0}) for matrix {1}".format(index, matrix_in))
     return r[0]['res']
-# ------------------------------------------------------------------------------
 
 # ------------------------------------------------------------------------------
 # -- Extreme value operations --------------------------------------------------
@@ -1361,7 +1356,6 @@ def matrix_min(schema_madlib, matrix_in, in_args, dim, matrix_out,
                fetch_index=False):
     matrix_extremum(schema_madlib, matrix_in, in_args, dim,
                     matrix_out, 'min', fetch_index)
-# ------------------------------------------------------------------------------
 
 # ------------------------------------------------------------------------------
 # -- Reduction operations ------------------------------------------------------
@@ -1524,7 +1518,6 @@ def matrix_mean(schema_madlib, matrix_in, in_args, dim):
                   'dense_agg_op': 'avg',
                   'dense_array_op': 'array_mean'}
     return matrix_reduction_op(schema_madlib, matrix_in, in_args, dim, vector_ops)
-# ------------------------------------------------------------------------------
 
 # ------------------------------------------------------------------------------
 # -- Mathematical operations ------------------------------------------------------