WIP: Use return type deduction to support lambda function #227

sdebionne · 2019-01-29T10:20:48Z

Allows to use gil variant (e.g. any_image<> ) visitation with generic lambdas:

gil::apply_operation(any_image_view<...>, [](auto view) {
   // do something with the view
});

Tasklist

Add test case(s)
Review
Adjust for comments
All CI builds and checks have passed

mloskot · 2019-01-29T10:31:21Z

include/boost/gil/extension/dynamic_image/apply_operation_base.hpp

+#define GIL_GENERATE_APPLY_FWD_OPS(N) BOOST_PP_REPEAT(N, GIL_APPLY_FWD_OP, BOOST_PP_EMPTY)
+
+namespace detail {
+	template <std::size_t N> struct apply_operation_fwd_fn {};


Do not increases the indentation level within namespace.
https://github.com/boostorg/gil/blob/develop/CONTRIBUTING.md#guidelines

mloskot · 2019-01-29T10:33:07Z

include/boost/gil/extension/dynamic_image/apply_operation_base.hpp

+
+// unary application
+template <typename Types, typename Bits, typename Op>
+auto BOOST_FORCEINLINE apply_operation_basec(const Bits& bits, std::size_t index, Op op) {


Could we try to keep the lines shorter, this way?

BOOST_FORCEINLINE auto apply_operation_basec(const Bits& bits, std::size_t index, Op op) { ... }

(@sdebionne I've just updated the snippet above)

BTW, there have been some discussion about possibility to get rid of BOOST_FORCEINLINE one day.

mloskot · 2019-01-29T10:33:26Z

include/boost/gil/extension/dynamic_image/apply_operation_base.hpp

+}
+
+namespace detail {
+	template <typename T2, typename Op>


Do not increases the indentation level within namespace.
https://github.com/boostorg/gil/blob/develop/CONTRIBUTING.md#guidelines

mloskot · 2019-01-29T10:33:49Z

include/boost/gil/extension/dynamic_image/apply_operation_base.hpp

+
+// Binary application by applying on each dimension separately
+template <typename Types1, typename Types2, typename Bits1, typename Bits2, typename Op>
+static auto BOOST_FORCEINLINE apply_operation_base(const Bits1& bits1, std::size_t index1, const Bits2& bits2, std::size_t index2, Op op) {


Write your code to fit within 90 columns of text
https://github.com/boostorg/gil/blob/develop/CONTRIBUTING.md#guidelines

mloskot · 2019-01-29T10:36:33Z

include/boost/gil/extension/dynamic_image/apply_operation_base.hpp

+#else
+
+#define GIL_APPLY_FWD_OP(z, N, text)                                                                        \
+    template <> struct apply_operation_fwd_fn<BOOST_PP_ADD(N,1)> {                                      \


I once had a wish to get rid of BOOST_PP_* :)

mloskot · 2019-01-29T10:39:20Z

include/boost/gil/extension/dynamic_image/apply_operation_base.hpp

+        }                                                                                               \
+        template <typename Types, typename Bits, typename UnaryOp>                                     \
+        auto applyc(const Bits& bits, std::size_t index, UnaryOp op) const { \
+            typedef typename mpl::begin<Types>::type                                             \


Would it be possible to replace all the MPL stuff with MP11 (http://boost.org/libs/mp11/) ?

This is a new code and it would help if it does not introduce any more uses of MPL which I'm trying to get rid of.

Even if MP11requires a bit of temporary plumbing e.g. to map boost::{true|false}_ to std::{true|false}_type, I think it is worth to try.

I can try but this is not really new code, it's mostly copy-and-paste and getting rid of the explicit return type, letting C++14 return type deduction do the work when available.

Right. I mean, it's worth a try, but I don't think it's a show-stopper, if you don't.

sdebionne · 2019-02-04T08:20:14Z

I'll fix the formatting issues (actually I think I have a less redondant solution) but I have a question regarding the CI: are there any tests that run with std=C++14? My modifications are only active when BOOST_NO_CXX14_RETURN_TYPE_DEDUCTION is set and hence depends on that compile options.

mloskot · 2019-02-04T10:41:50Z

@sdebionne

Are there any tests that run with std=C++14?

No, we currently don't have any builds for C++14 mode.

I think it would be a good idea to add one, and for C++17 too.

Do you have any preference regarding CI service?

Do you need a specific compiler version? See table in the README for what we've got on which CI. On CircleCI, we have these available without any extra installation required:

gil/.circleci/config.yml

Lines 15 to 75 in 7722f08

    
           docker_gcc48: &docker_gcc48 
        
             docker: 
        
               - image: gcc:4.8 
        
           docker_gcc49: &docker_gcc49 
        
             docker: 
        
               - image: gcc:4.9 
        
           docker_gcc51: &docker_gcc51 
        
             docker: 
        
               - image: gcc:5.1 
        
           docker_gcc52: &docker_gcc52 
        
             docker: 
        
               - image: gcc:5.2 
        
           docker_gcc53: &docker_gcc53 
        
             docker: 
        
               - image: gcc:5.3 
        
           docker_gcc54: &docker_gcc54 
        
             docker: 
        
               - image: gcc:5.4 
        
           docker_gcc55: &docker_gcc55 
        
             docker: 
        
               - image: gcc:5.5 
        
           docker_gcc61: &docker_gcc61 
        
             docker: 
        
               - image: gcc:6.1 
        
           docker_gcc62: &docker_gcc62 
        
             docker: 
        
               - image: gcc:6.2 
        
           docker_gcc63: &docker_gcc63 
        
             docker: 
        
               - image: gcc:6.3 
        
           docker_gcc64: &docker_gcc64 
        
             docker: 
        
               - image: gcc:6.4 
        
           docker_gcc71: &docker_gcc71 
        
             docker: 
        
               - image: gcc:7.3 
        
           docker_gcc72: &docker_gcc72 
        
             docker: 
        
               - image: gcc:7.2 
        
           docker_gcc73: &docker_gcc73 
        
             docker: 
        
               - image: gcc:7.3 
        
           docker_gcc82: &docker_gcc82 
        
             docker: 
        
               - image: gcc:8.2 
        
           docker_clang39: &docker_clang39 
        
             docker: 
        
               - image: rferraro/cxx-clang:3.9 
        
           docker_clang40: &docker_clang40 
        
             docker: 
        
               - image: rferraro/cxx-clang:4.0 
        
           docker_clang50: &docker_clang50 
        
             docker: 
        
               - image: rferraro/cxx-clang:5.0

Keep in mind, Travis CI usually starts building with looong delays.

Let me know and I will add new build job, then you can update your PR.

sdebionne · 2019-02-05T08:30:59Z

Superseded by #231.

@mloskot

Do you have any preference regarding CI service?

Nope, whatever is easier for you.

Do you need a specific compiler version? See table in the README for what we've got on which CI. On CircleCI, we have these available without any extra installation required:

According to compiler support, anything above gcc 5 is fully c++14 compliant. Personally I use gcc 7.3 as it's the one that ship with Conda but that is not a requirement.

Azure pipelines look nice and reactive.

mloskot · 2019-02-05T08:41:19Z

OK, I will try to work it out today.

Use return type deduction to support lambda function

f22f0f4

mloskot requested changes Jan 29, 2019

View reviewed changes

mloskot added ext/dynamic_image boost/gil/extension/dynamic_image cat/enhancement Improvements, but not fixes addressing identified bugs labels Jan 29, 2019

mloskot requested a review from stefanseefeld January 29, 2019 12:05

mloskot added the status/work-in-progress Do NOT merge yet until this label has been removed! label Jan 29, 2019

sdebionne mentioned this pull request Feb 4, 2019

Use Boost.Variant instead of GIL's own variant implementation. #231

Merged

3 tasks

sdebionne closed this Feb 5, 2019

mloskot added status/duplicate and removed status/work-in-progress Do NOT merge yet until this label has been removed! labels Feb 5, 2019

mloskot mentioned this pull request Feb 5, 2019

Add C++14 builds using GCC 8 and VS2017 to Azure Pipelines #233

Merged

1 task

sdebionne deleted the apply-operation-lambda branch October 20, 2020 16:40

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

WIP: Use return type deduction to support lambda function #227

WIP: Use return type deduction to support lambda function #227

sdebionne commented Jan 29, 2019

mloskot Jan 29, 2019

mloskot Jan 29, 2019 •

edited

Loading

mloskot Jan 29, 2019

mloskot Jan 29, 2019

mloskot Jan 29, 2019

mloskot Jan 29, 2019

sdebionne Jan 29, 2019

mloskot Jan 29, 2019

sdebionne commented Feb 4, 2019

mloskot commented Feb 4, 2019

sdebionne commented Feb 5, 2019

mloskot commented Feb 5, 2019

WIP: Use return type deduction to support lambda function #227

WIP: Use return type deduction to support lambda function #227

Conversation

sdebionne commented Jan 29, 2019

Tasklist

mloskot Jan 29, 2019

Choose a reason for hiding this comment

mloskot Jan 29, 2019 • edited Loading

Choose a reason for hiding this comment

mloskot Jan 29, 2019

Choose a reason for hiding this comment

mloskot Jan 29, 2019

Choose a reason for hiding this comment

mloskot Jan 29, 2019

Choose a reason for hiding this comment

mloskot Jan 29, 2019

Choose a reason for hiding this comment

sdebionne Jan 29, 2019

Choose a reason for hiding this comment

mloskot Jan 29, 2019

Choose a reason for hiding this comment

sdebionne commented Feb 4, 2019

mloskot commented Feb 4, 2019

sdebionne commented Feb 5, 2019

mloskot commented Feb 5, 2019

mloskot Jan 29, 2019 •

edited

Loading