Hi there,

I'm trying to build polygons from a list of vertices. However, there seems no S2Builder port to pywraps2. How can we build polygons effectively using the Python port?

I've actually tried to build it with S2Loop and S2Polygon,

loop = S2Loop()
points = [S2LatLng.FromDegrees(float(vertex[0]), float(vertex[1])).ToPoint() for vertex in hull[:-1]]
loop.Init(points)
loop.Normalize()
polygon = S2Polygon(loop)

The problem is that S2Loop requires a lot of checking, like edges not intersecting each other. And right now there are no easy ways for me to ensure the input (list of vertices) would not cross with each other (the input is from a concave hull calculation), so I often ran into errors like ERROR Edge 46 crosses edge 48.

Are there any effective ways to build polygons in Python given the circumstance above?

Thanks a lot,
Foxan

Hi *

I would need the functionality/method to get all S2Cells for a given (very precise) polygon, lets say a country, that returns all S2Cells outside the polygon, i.e., the 5 faces plus all other S2Cells of the polygon surrounding the given area in face 6.

Is that somehow possible to calculate?

Kind regards + thx for any reply
Sebastian

First off, thanks for the great library! It has been a joy to use!

My issue is that it was difficult getting S2 to work with my project due to it not being namespaced. Most primitive classes, typedefs, and structs in the S2 library (e.g. S2Polygon, S2CellId, S2Point), as well as some classes with common names in //utils (e.g. Varint, Vector3_d) are declared in the global namespace. These can lead to conflicts, or even if they don't, should generally not be in a global namespace. Similarly for many of the macros and functions in //base (especially //base/port.h), or for //base/mutex.h which appears to collide with any target including its own absl/synchronization/mutex.h.

It would be nice if all code was wrapped in namespace s2 or similar, and all global macros had an S2_ prefix.

Would there be interest in creating a node.js wrapper with SWIG? I know it would be helpful to me as I mostly work in node.js and node-s2 from mapbox is getting out-of-date.

PS. I'd be willing to help create this. However, since I'm unfamiliar with SWIG and building node.js C++ interfaces more generally, it's probably not best that I lead.

I have two polygons:

1. First (A) is a circle of radius R

2. Second (B) is a circle of radius less than R (for example R/2)
   Both circles has same center. Now im trying to InitToDifference C = A - B, so the right answer, i guess, is polygon that consist of A and reverse B, but the result is {A, B}.
   There is code without generation of A and B. Is it a bag or am i wrong?

   S2Polygon* union_res = new S2Polygon;
   std::cout << "First:\n";
   for(int i = 0; i < a->num_loops(); i++) {
   std::cout << a->loop(i)->GetArea() << "\n";
   }

   std::cout << "Second:\n";
   for(int i = 0; i < b->num_loops(); i++) {
   std::cout << b->loop(i)->GetArea() << "\n";
   }
   union_res->InitToDifference(a, b);
   std::cout << "Out:\n";
   for(int i = 0; i < union_res->num_loops(); i++) {
   std::cout << union_res->loop(i)->GetArea() << "\n";
   }

Output:
First:
0.00829343
Second:
0.0022473
Out:
0.00829343
0.0022473

Hi, about pywraps2_test.py, I try python pywraps2_test.py but genereted a error: "ImportError: No module named pywraps2".

How to install pywraps2 module? Perhaps the README.md need some clues for dummy users.

NOTES: I am using UBUNTU 16 LTS and followed the building instructions and the installation was fine, with make tests sucessess. SWIG installation also fine.

Is S2PointIndex thread safe? Should I use mutex when S2PointIndex.Add() and S2ClosestPointQuery.query() are called simultaneously in multithread?

Hi,

when trying to create objects of S2Polyline or S2Polgon I get the following error:

undefined reference to 'S2Polyline::S2Polyline()' undefined reference to 'S2Polyline::Init(std::vector<S2LatLng, std::allocator<S2LatLng> > const&)' undefined reference to 'S2Polyline::~S2Polyline()' undefined reference to 'S2Polyline::~S2Polyline()'

The code I am using is simply this:
std::vector<S2LatLng> vertices; S2Polyline polyline; polyline.Init(vertices);

The implementation of the declared functions in the header files (s2polyline.h and s2polygon.h) seems to be missing.

Thank you in advance
Best regards

here my Cmakelists is:
`project(test)

set(CMAKE_CXX_STANDARD 17)

set(INC_DIR /usr/local/include/s2)
set(LINK_DIR /usr/local/lib)

include_directories(${INC_DIR})
link_directories(${LINK_DIR})
link_libraries(libs2.so)
here is the error:
'add_executable(test main.cpp) undefined reference to S2Testing::RandomPoint()' undefined reference to S2Testing::RandomPoint()'`
I built this code yesterday. How can i do?Thanks!

If DCHECK is disabled, the code inside it will not run, causing the function to give a wrong result.

Could anyone tell me how to integrate s2geometry into geomesa?Thanks!
https://www.geomesa.org/documentation/tutorials/geohash-substrings.html

I am looking for a S2 geometry version of this SE answer - https://gis.stackexchange.com/questions/18330/using-geohash-for-proximity-searches/92331#92331

1. I want to use S2 cells to do approximate nearest neighbor searches

2. I have a bounding box on a 2D grid i.e. composed of latitudes and longitudes. For each latitude and longitude I want a S2 cell.These need to be "indexed" with a particular precision. That precision needs to be linked to a radius in kilometers or meters.

3. Then I have a whole bunch of query coordinates completely unrelated to the above coordinates. So I can convert this coordinate to a index as well and maybe get's neighbors as well. These can be indexed at a lower precision.

4. Then use this index to return nearest neighbors from the original bounding box. Here I am assuming some bit shifting will have to be done.

I have a highly optimized SkipList which does fast inserts and range queries. So the range query would return coordinates from the bounding box given the lower and upper range values from (3).

Is that possible with s2 geometry ?

Who can tell me:
1.What's the advantage and disadavntage of s2? What scenes could s2 to be used?
2.Compare to R-Tree, what's the advantage and disadvantage of s2?
Thanks!

The Readme says

Requirements for End Users

• ...
• googletest testing framework (to build tests and example programs, optional)

Despite this, the s2testing library is included in the build unconditionally, which in turn includes the file s2shapeutil_testing.cc, which includes gtest/gtest.h.

So it seems to me that googletest is required in practice. I'm not sure if the readme just needs to be updated to say that googletest is required or if the s2testing library should be included in the conditional test block.

I find the CMake file is a bit out of date and I would like to rewrite it to make it easier to use. In preparation for that, I would like to bump the version to 3.5.
I think 3.5 is the lowest version it should be since Ubuntu 16.04 comes with 3.5. I don't think this should inconvenience people too much. Later versions of CMake are very easy to install and Ubuntu 18.04 comes with CMake 3.10.

Given lat/lon, how do I get the cell ID at a certain level? I see a constructor for lat/lon only, which as far as I understand provides the cell ID at the maximum level only.

Hello, friends,

I found in the s2boolean_operation.h file, the author mentioned S2MultiBooleanOperation which can do boolean operations on multiple objects. However I cannot find any where in the code that has this S2MultiBooleanOperation class. Any ideas why?

I need to union a set of S2Polygons and I find using S2BooleanOperation to do this union one by one to be very slow.

If anyone could give me some advise, that would be great! Thank you.

Best,

Hi,

I'm trying to encode and decode a polygon using as an example the C++ tests, and the Encoder() class doesn't seem to work the same, or maybe it isn't yet exposed in the SWIG files.

[GCC 4.2.1 Compatible Apple LLVM 8.0.0 (clang-800.0.34)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import pywraps2 as s2
>>> london = s2.S2LatLng.FromDegrees(51.5001525, -0.1262355)
>>> src = s2.S2Polygon(s2.S2Cell(s2.S2CellId(london)))
>>> encoder = s2.Encoder()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'module' object has no attribute 'Encoder'

Is there a different way to accomplish this in python?

I have polygon CW POLYGON ((-99 3, 92 3, 92 29, -99 29))
which return FULL Bound [-180:180,-90:90].
Is this correct ?

third_party/absl/ will conflict abseil-cpp

I feel a little silly asking this, but is there some kind of magic required to build the test programs? I figured it was a simple way to check my installation of the library, but I cannot even get past the includes :)

$ cc -I/usr/local/include -I/opt/local/include -o term_index term_index.cc 
In file included from term_index.cc:21:
In file included from /usr/local/include/s2/s2earth.h:25:
In file included from /usr/local/include/s2/s1angle.h:28:
In file included from /usr/local/include/s2/s2point.h:22:
In file included from /usr/local/include/s2/util/math/vector.h:32:
In file included from /usr/local/include/s2/base/logging.h:54:
In file included from /usr/local/include/s2/base/log_severity.h:25:
/usr/local/include/s2/third_party/absl/base/log_severity.h:32:12: error: ISO C++ forbids forward references to 'enum' types
enum class LogSeverity : int {
           ^
/usr/local/include/s2/third_party/absl/base/log_severity.h:32:24: error: expected unqualified-id
enum class LogSeverity : int {
                       ^
/usr/local/include/s2/third_party/absl/base/log_severity.h:41:1: error: unknown type name 'constexpr'
constexpr std::array<absl::LogSeverity, 4> LogSeverities() {
^
/usr/local/include/s2/third_party/absl/base/log_severity.h:41:28: error: no member named 'LogSeverity' in namespace 'absl'
constexpr std::array<absl::LogSeverity, 4> LogSeverities() {
                     ~~~~~~^
/usr/local/include/s2/third_party/absl/base/log_severity.h:41:44: error: cannot define or redeclare 'LogSeverities' here
      because namespace 'absl' does not enclose namespace 'std'
constexpr std::array<absl::LogSeverity, 4> LogSeverities() {

Hi

Does the Python API support all native methods?

I would need set_min_level, set_max_level, and set_max_cells in S2RegionCoverer.

Or is it possible to access native methods from within python?

Kind regards

Thank you for your contribution.I use S2RegionCoverer class to get specific region covering cells,and i get overlaps among adjacent google s2 cells.Here is my code.
` S2RegionCoverer::Options options;
options.set_min_level(level);
options.set_max_level(level);
S2RegionCoverer s2RegionCoverer(options);

// s2RegionCoverer.GetInteriorCovering(region,interior);
s2RegionCoverer->GetCovering(region,interior);I try three type ofregion,S2LatLngRect,S2Loop,S2Polygon,and the result is same.Could you give me some Suggest?

The file str_cat.h seems to be missing from installation in CMakelists.txt, causing failure when compiling with some of the s2 header files:

#include <s2/s2point_index.h>

int main() {
}  

Compiling this file gives an error:

$ g++ -std=c++11 test.cc 
In file included from /usr/local/include/s2/util/gtl/btree.h:74:0,
                 from /usr/local/include/s2/util/gtl/btree_map.h:34,
                 from /usr/local/include/s2/s2point_index.h:24,
                 from test.cc:1:
/usr/local/include/s2/util/gtl/layout.h:183:10: fatal error: s2/third_party/absl/strings/str_cat.h: No such file or directory
 #include "s2/third_party/absl/strings/str_cat.h"
          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
$ ls /usr/local/include/s2/third_party/absl/strings/
string_view.h
$

Does S2 support efficient overlap querying? I want to index polygons which can vary in size a lot (from room size to continent size), and efficiently check whether a point is within a region.
A nearby search would find me the nearby vertices only, but not those covering the point which could be far away.

I'm trying to deal with feature to eliminate self-intersections. As written in s2builder.h

// - It can eliminate self-intersections by finding all edge pairs that cross
// and adding a new vertex at each intersection point.

and

// - Guaranteed minimum separation between edges and vertices other than
// their endpoints (similar to the goals of Iterated Snap Rounding). In
// other words, edges that do not intersect in the output are guaranteed
// to have a minimum separation between them."

Therefore when I use a builder I expect that output will meet the requrements to data for s2loop. But edges intersect point appears in output twice. That data is incorrect for s2loop.
I have checked a property of SnapFunction "min_edge_vertex_separation" and it is non zero.

// 5. A "min_edge_vertex_separation", the guaranteed minimum distance
// between edges and non-incident vertices in the output. This is
// generally a fraction of "snap_radius" where the fraction depends on
// the snap function.

In example below I'm trying to build s2loop from square with crossing diagonals. Here are many excess options, but i tried them all to be sure.

https://pastebin.com/WBpgn6Vj

This code crashs on S2loop::Init

/home/noscoper/s2geometry-master/src/s2/s2loop.cc:134 ERROR Edge 1 has duplicate vertex with edge 4
/home/noscoper/s2geometry-master/src/s2/s2loop.cc:287 FATAL Check failed: IsValid() Signal: SIGABRT (Aborted)

What is wrong? Is it my missunderstood or bug?

FindClosestCellsBruteForce()
FindClosestCellsOptimized()
will conduct different results

issue errors on s2closest_cell_query_test at TestFindClosestCells(...)

I was looking at the code, and it'd be nice if I could check if S2Loop was in an S2Cap. Naively, it seems like you can take the existing code for S2Cap::Contains(const S2Cell& cell) and S2Cap::Intersects(const S2Cell& cell, ...) and just generalize it for an S2Loop, at a minimum, and maybe S2Rect and S2Poly as well. Would this be reasonable? Or is there a better way at handling this?

Obviously an arbitrary S2Region Intersects/Contains S2Region would be ideal, but I wasn't sure if that made sense to the other classes that extend S2Region.

I noticed quite a lot of warnings in clang due to signed / unsigned integer comparison as well as structs that were forward declared as classes.

Would it be ok to submit a pull request which silences some of these ?

Hi,

I get the following error when compiling code that is using s2latlng:

In file included from /usr/local/include/s2/base/logging.h:54:0, from /usr/local/include/s2/util/math/vector.h:32, from /usr/local/include/s2/r2.h:22, from /usr/local/include/s2/s2latlng.h:28, ... (removed own files here) /usr/local/include/s2/base/log_severity.h:25:51: fatal error: s2/third_party/absl/base/log_severity.h: No such file or directory compilation terminated.

I think this has to to with the commit of March 5th (Make gflags and glog optional.). Maybe the cmake install configuration has to be adapted?
Using the commit of Feb 27 it worked fine.

If there is something that I am doing wrong pls let me know.

Thank you in advance
Best regards

https://github.com/google/s2geometry/blame/master/CMakeLists.txt#L61

You should add this flag for clang/gcc to the private target definitions only. Or better use set( CMAKE_CXX_STANDAD 11) / comple features . It could look like: target_compile_features(s2 PRIVATE cxx_std_11). Please use something newer than cmake 2.8.x. You have Blaze internally so don't force people to use antique cmake versions.

Hi

What would be the easiest way to get from a S2CellUnion to MultiPolygons which can be used in GeoJSON or QGIS etc. to get a visual representation of the given S2CellUnion and overlay it with a city polygon for example?

Thx for any help

Hi, I am fortunate to be able to use this S2 algorithm. I have to say that this is a great project. I admit I underestimated it. But when I use the swig python version, I can not use the method I want, pywraps2.S2RegionCoverer (), I can not use the set_min_level (), set_max_level (), set_max_cells () I was expecting because I found my project file Without these, I do not know where the problem is. If you can help me, I will be grateful.

Hello, first of all thanks for the great lib! I have been using the Go version successfully before but moving to C++ since I need S2ClosestEdgeQuery which is not ported yet.

However, I am having a strange issue when trying to create a S2CrossingEdgeQuery. I am using the following code:

int main(int argc, char* argv[]) {
    const auto& geo = readGeoJSON(argv[1], false);

    MutableS2ShapeIndex index;

    const auto& features = geo.get<feature_collection>();
    for (auto f : features) {
        std::vector<S2Point> pVector;
        const auto& p = f.geometry.get<polygon>();
        const auto& ring = p.at(0);
        for (auto coord: ring) {
            S2Point p = S2LatLng::FromDegrees(coord.y, coord.x).ToPoint().Normalize();
            pVector.push_back(p); 
        }
        pVector.pop_back();
        S2Loop loop(pVector);
        auto loopShape = absl::make_unique<S2Loop::Shape>(&loop);
        index.Add(absl::move(loopShape));
    }
    
    std::cout << "Create closest edge q" << std::endl;
    auto qtest = new S2ClosestEdgeQuery(&index);
    
    std::cout << "Create crossing edge q" << std::endl;
    auto query = new S2CrossingEdgeQuery(&index);
    std::cout << "Done!" << std::endl;
}

The readGeoJSON function will read the following GeoJSON:

{
  "type": "FeatureCollection",
  "features": [
    {
      "type": "Feature",
      "properties": {},
      "geometry": {
        "type": "Polygon",
        "coordinates": [
          [
            [
              11.39007568359375,
              57.787696727659075
            ],
            [
              11.217041015625,
              57.676254483720356
            ],
            [
              11.483459472656248,
              57.66156546307091
            ],
            [
              11.53289794921875,
              57.81989228369527
            ],
            [
              11.39007568359375,
              57.787696727659075
            ]
          ]
        ]
      }
    }
  ]
}

To debug, I added the following to s2edge_crosser.h:257:

inline void S2EdgeCrosser::RestartAt(const S2Point* c) {
  S2_LOG(INFO) << *c;
  ...
}

What happens is that I see a bunch of logs like:

/s2geometry/src/s2/s2edge_crosser.h:258 INFO [0.52256, 0.105272, 0.846079]

however, when it gets to creating the S2CrossingEdgeQuery I see the following:

Create crossing edge q
/s2geometry/src/s2/s2edge_crosser.h:258 INFO [0, 0.105272, 0.846079]
/s2geometry/src/s2/s2edge_crosser.h:259 FATAL Check failed: S2::IsUnitLength(*c) Aborted

It seems like the first coordinate of the point is suddenly zero?
I'm sure this might be some issue on my part, but any help is greatly appreciated!

Hi,

I was trying to compile s2 python binding and had to update CMakefile a bit as I'm using newer version of CMake and some of directives are deprecated:

https://gist.github.com/hankim/d058ba1f26d9efd5aba296d6fa28e2e3

How to cover specified polyline or polygon with specified level cell?Thanks!

Adding the following line to CMakeLists.txt

target_include_directories(s2 PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/src)

would allow to seamlessly include S2 in a different CMake project like this:

list(APPEND CMAKE_MODULE_PATH "<path_to_s2geometry_folder>/third_party/cmake")
add_subdirectory(<path_to_s2geometry_folder> s2geometry)
target_link_libraries(<target_name> s2)

Hi, I am trying to initialize an S2Polygon from a bunch of loops by calling Normalize on each loop and then calling InitNested. The polygon is for San Jose, California. For some reason the polygon is coming out inverted, covering the whole Earth except San Jose, even though all of its constituent loops are normalized. Can you help?

Note: in all of my loops, the last vertex is the same as the first vertex. If I explicitly remove that identical last vertex, the issue goes away. I had thought that S2 would be OK with this imperfect data. Is there any functionality in S2 that I can use to not have to prune those vertices myself?

Here is a C++ program to replicate: https://paste.ee/p/7FAdB

as of 8a8746c

compilation is failing on Arch because of openssl 1.1:

In file included from /home/akh/tmp/s2geometry/src/s2/s2predicates_internal.h:29:0,
                 from /home/akh/tmp/s2geometry/src/s2/s2edge_crossings.cc:28:
/home/akh/tmp/s2geometry/src/s2/util/math/exactfloat/exactfloat.h:516:10: error: field ‘bn_’ has incomplete type ‘BIGNUM {aka bignum_st}’
   BIGNUM bn_;
          ^~~
In file included from /usr/include/openssl/bn.h:32:0,
                 from /home/akh/tmp/s2geometry/src/s2/util/math/exactfloat/exactfloat.h:119,
                 from /home/akh/tmp/s2geometry/src/s2/s2predicates_internal.h:29,
                 from /home/akh/tmp/s2geometry/src/s2/s2edge_crossings.cc:28:
/usr/include/openssl/ossl_typ.h:80:16: note: forward declaration of ‘BIGNUM {aka struct bignum_st}’
 typedef struct bignum_st BIGNUM;
                ^~~~~~~~~
In file included from /home/akh/tmp/s2geometry/src/s2/s2predicates_internal.h:29:0,
                 from /home/akh/tmp/s2geometry/src/s2/s2edge_crossings.cc:28:
/home/akh/tmp/s2geometry/src/s2/util/math/exactfloat/exactfloat.h: In constructor ‘ExactFloat::ExactFloat()’:
/home/akh/tmp/s2geometry/src/s2/util/math/exactfloat/exactfloat.h:576:3: error: ‘BN_init’ was not declared in this scope
   BN_init(&bn_);
   ^~~~~~~
/home/akh/tmp/s2geometry/src/s2/util/math/exactfloat/exactfloat.h:576:3: note: suggested alternative: ‘BN_print’
   BN_init(&bn_);
   ^~~~~~~
   BN_print
make[2]: *** [CMakeFiles/s2.dir/build.make:879: CMakeFiles/s2.dir/src/s2/s2edge_crossings.cc.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:105: CMakeFiles/s2.dir/all] Error 2
make: *** [Makefile:130: all] Error 2

BN_init() was removed in OpenSSL 1.1.0; use BN_new() instead.
source https://www.openssl.org/docs/man1.1.0/crypto/BN_new.html

Solution for Arch is:
cmake -DOPENSSL_INCLUDE_DIR="/usr/include/openssl-1.0" -DOPENSSL_SSL_LIBRARY=/usr/lib/openssl-1.0/libssl.so ..

It will become an issue for every distribution with recent openssl.

We will need to do something about what are currently external dependencies: openssl, glog, gflags.

This might be useful:
openssl/openssl#3840

C:\Users\shkil\Downloads\s2geometry-master\s2geometry-master\src\s2/third_party/absl/container/internal/layout.h(352): error C2971: std::tuple_element:

Hi, great to see the new version of S2 out - very cool!

I know a lot of people use the old version of S2 (us for example) and link it statically into their application. That was fairly easy to do with the old version, but now with the GFlags, GLog dependencies it's a lot more onerous and also affects downstream build system requirements.

Is there any chance you might consider an option to allow building the core without requiring any dependencies, aside from libssl of course.

Best

I got 5 errors on MacOS running make,

is there maybe

#include

missing in s2region.h and s2cap.h?

That at least seemed to solve the issues.

The current cmake configuration does not search for the correct python site-packages directory and install the bindings in PREFIX/share/python, so it can't be loaded by the interpreter as is.

One solution would be to look for site-packages directory via python -c "from distutils.sysconfig import get_python_lib; print(get_python_lib())"

The right way is probably to use python itself to build & install the bindings.

Hey, having read #40, I'd like to push again for a tagged release. Reason for asking is that package managers, e.g. Homebrew, won't accept a package for S2 unless there is a tagged version. It would be very much appreciated.

I followed the instructions as given here - http://s2geometry.io/about/platforms.html but I am getting these errors when I run cmake

-- The following features have been enabled:

• GFLAGS, allows changing command line flags.
• GLOG, provides logging configurability.
• SHARED_LIBS, builds shared libraries instead of static.

GTEST_ROOT: /usr/src/googletest
File "/usr/lib/python3.6/site.py", line 183
file=sys.stderr)
^
SyntaxError: invalid syntax
CMake Error at src/python/CMakeLists.txt:30 (install):
install TARGETS given no LIBRARY DESTINATION for module target "_pywraps2".

CMake Error at src/python/CMakeLists.txt:31 (install):
install FILES given no DESTINATION!

-- Configuring incomplete, errors occurred!
See also "/home/winash12/S2source/s2geometry/CMakeFiles/CMakeOutput.log".
See also "/home/winash12/S2source/s2geometry/CMakeFiles/CMakeError.log".
This is on Ubuntu 18.04 LTS.

All the links here are broken; e.g. this one. They were working a day or two ago.

I want to use s2geometry as a thirdparty library in my project, and don't want to modify it frequently, I find that the zip (https://github.com/google/s2geometry/archive/master.zip) is always refer to the latest master branch, so i want know is there any stable release version?
thanks

I cloned and built the latest master branch.
I could find the MayIntersect method in the S2Polyline object however it is in the CPP source code.

Help needed to figure out the way to call this method in Python.

Thanks a lot.

After compile and install s2, In the s2geometry/doc/examples/ directory. I run the following command
g++ term_index.cc -std=c++17 -I/usr/local/lib/ -ls2
it throw out link error:
/tmp/ccxc1b2Q.o：in ‘main’：
term_index.cc:(.text+0x5a)：undefined reference to ‘S2Testing::RandomPoint()’
term_index.cc:(.text+0x2b5)：undefined reference to ‘S2Testing::RandomPoint()’
collect2: error: ld returned 1 exit status
Actually, I have included the s2testing.h yet.
#include "s2/s2testing.h"