Updated docs for modern CMake #124

Author: serizba

docs/source/examples.rst

@@ -1,19 +1,6 @@
 Examples
 ========
 
-This guide show how to create and run some basic examples that show the functionality of cppflow. The code of these examples is available on the `GitHub repo <https://github.com/serizba/cppflow/tree/cppflow2/examples>`_. To run these examples, just use the provided CMake on each of them (after having installed the TF C API as in :ref:`Installation`):
-
-.. code:: bash
-
-    git clone git@github.com:serizba/cppflow.git
-    cd cppflow/examples/load_model
-    mkdir build
-    cd build
-    cmake ..
-    make
-    ./example
-
-
 Create and load model
 ---------------------
 

docs/source/installation.rst

@@ -3,7 +3,7 @@
 Installation
 ============
 
-One of the advantages of cppflow is that you don't need to compile or install TensorFlow, you just need to download the `TF C API <https://www.tensorflow.org/install/lang_c>`_. As cppflow is a header-only library, once you have the C API, the only thing you need to do is include the cppflow files from your project.
+One of the advantages of cppflow is that you don't need to compile or install TensorFlow, you just need to download the `TF C API <https://www.tensorflow.org/install/lang_c>`_. Cppflow is a header-only library, and thus you can just include the cppflow files in your project.
 
 To install the C API in your system you have two options:
 
@@ -18,12 +18,8 @@ You can install the C API in a system directory and do not worry about it again.
 
 Install the TF C API in custom directory
 ----------------------------------------
-.. note::
-    The easiest way is to extract the library into your HOME directory, in a folder called ``libtensorflow2``. In this case, you can directly use the CMake files from the examples, which will search for the library in your HOME directory.
-
-    ``mkdir -p ~/libtensorflow2/ && tar -C ~/libtensorflow2/ -xzf (downloaded file)``
 
-You can also install the library in a custom directory. In this case, after `downloading it <https://www.tensorflow.org/install/lang_c>`_ and unpacking it you will need to update your PATH or add a hint to your ``CMakeLists.txt`` to find the library.
+You can also install the library in a custom directory. In this case, after `downloading it <https://www.tensorflow.org/install/lang_c>`_ and unpacking it you will need to update your PATH or tell CMake where you placed the library with ``-DCMAKE_PREFIX_PATH=...``.
 
 .. code:: bash
 
@@ -37,11 +33,24 @@ Now, update your path:
     export LIBRARY_PATH=$LIBRARY_PATH:/path/to/mydir/lib
     export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/mydir/lib
 
-Or add a hint in ``CMakeLists.txt`` as in the examples:
 
-.. code:: cmake
+Install cppflow
+-----------------------------
+
+Cppflow is just a header-only library, and thus it does not require to build it. To facilitate the installation, we provide a CMake file that will install the library in your system. To install it, you just have to:
+
+.. code:: bash
+
+    mkdir build
+    cd build
+    cmake ..
+    make -j
+    make install
+
+.. note::
+    If you installed the TF C API in a custom directory, you will need to tell CMake where you placed the library with ``-DCMAKE_PREFIX_PATH=/path/to/mydir/``.
 
-    find_library(TENSORFLOW_LIB tensorflow HINT /path/to/mydir/lib)
+This will also compile the examples, if you don't want to compile them, you can use ``-DBUILD_EXAMPLES=OFF``.
 
 
 You are done, now you can proceed to build your :ref:`first example<First example>`.

docs/source/quickstart.rst

@@ -12,7 +12,7 @@ Once you have downloaded the TF C API and cppflow you can start playing with ten
 .. code:: c++
 
     #include <iostream>
-    #include "cppflow/cppflow.h"
+    #include <cppflow/cppflow.h>
 
     int main() {
 
@@ -26,11 +26,11 @@ Once you have downloaded the TF C API and cppflow you can start playing with ten
         return 0;
     }
 
-Easy right?, now you can compile it with the terminal (if you have configured the TF C API as stated in :ref:`Installation`) using the following command:
+Easy right?, now you can compile it with the terminal (if you have configured the TF C API and installed cppflow as stated in :ref:`Installation`) using the following command:
 
 .. code:: bash
 
-    g++ -std=c++17 -o main.out -I /path/to/libtensorflow2/include/ -I/path/to/cppflow/include/ main.cpp -ltensorflow
+    g++ -std=c++17 -o main.out main.cpp -ltensorflow
     ./main.out
 
 You should see the result of ``a + b``:
@@ -48,17 +48,34 @@ Probably a more convenient way of compiling your code is using CMake.
 .. code:: cmake
 
     cmake_minimum_required(VERSION 3.10)
+
     project(example)
 
-    find_library(TENSORFLOW_LIB tensorflow HINT $ENV{HOME}/libtensorflow2/lib)
+    add_executable(example main.cpp)
+
+    find_package(cppflow REQUIRED)
 
-    set(CMAKE_CXX_STANDARD 17)
+    target_include_directories(
+    example PUBLIC
+    cppflow::cppflow
+    )
 
-    add_executable(example main.cpp)
-    target_include_directories(example PRIVATE /path/to/cppflow/include/ $ENV{HOME}/libtensorflow2/include)
-    target_link_libraries (example "${TENSORFLOW_LIB}")
+    target_link_libraries(
+    example PUBLIC
+    cppflow::cppflow
+    )
 
+Now you can compile it with:
 
+.. code:: bash
+
+    mkdir build
+    cd build
+    cmake ..
+    make
+
+.. note::
+    If you installed the TF C API or cppflow in a custom directory, you will need to tell CMake where you placed them ``-DCMAKE_PREFIX_PATH=/path/to/mydir/``.
 
 Load a model
 ------------
@@ -68,7 +85,7 @@ You can easily run TensorFlow models with cppflow by loading a `saved model <htt
 .. code:: c++
 
     #include <iostream>
-    #include "cppflow/cppflow.h"
+    #include <cppflow/cppflow.h>
 
 
     int main() {

examples/eager_op_multithread/main.cpp

@@ -3,7 +3,7 @@
 #include <vector>
 #include <iostream>
 
-#include "cppflow/cppflow.h"
+#include <cppflow/cppflow.h>
 
 constexpr size_t num_iter = 10240;
 constexpr size_t num_threads = 32;

examples/efficientnet/main.cpp

@@ -1,6 +1,6 @@
 #include <iostream>
 
-#include "cppflow/cppflow.h"
+#include <cppflow/cppflow.h>
 
 
 int main() {

examples/load_frozen_graph/main.cpp

@@ -1,7 +1,7 @@
 #include <iostream>
 
-#include "cppflow/ops.h"
-#include "cppflow/model.h"
+#include <cppflow/ops.h>
+#include <cppflow/model.h>
 
 
 int main() {

examples/load_model/main.cpp

@@ -1,7 +1,7 @@
 #include <iostream>
 
-#include "cppflow/ops.h"
-#include "cppflow/model.h"
+#include <cppflow/ops.h>
+#include <cppflow/model.h>
 
 
 int main() {

examples/multi_input_output/main.cpp

@@ -1,7 +1,7 @@
 #include <iostream>
 
-#include "cppflow/ops.h"
-#include "cppflow/model.h"
+#include <cppflow/ops.h>
+#include <cppflow/model.h>
 
 int main() {
 

examples/tensor/main.cpp

@@ -3,7 +3,7 @@
 #include <iostream>
 #include <stdexcept>
 
-#include "cppflow/cppflow.h"
+#include <cppflow/cppflow.h>
 
 bool float_equal(const float f1, const float f2) {
     return std::abs(f1/f2-1.0f) < 1e-6;

examples/tensor/odr.cpp

@@ -1,4 +1,4 @@
-#include "cppflow/cppflow.h"
+#include <cppflow/cppflow.h>
 
 // Do NOT remove this file
 // test ODR violation