[readers.ept] resolution option strange behaviour

[gui2dev]

Hi,

The resolution option for readers.ept seems to behave in a strange way :

This program :

main.cpp

#include <pdal/PointTable.hpp>
#include <pdal/PointView.hpp>
#include <pdal/StageFactory.hpp>
#include <pdal/PluginManager.hpp>
#include <pdal/io/BufferReader.hpp>
#include <pdal/Options.hpp>
#include <pdal/filters/StreamCallbackFilter.hpp>

  using namespace pdal;

int main(int argc, char* argv[])
{
  std::unique_ptr<StageFactory> factory(new StageFactory);
  
Options readOpts;
  readOpts.add(Option("filename", "path/to/my/ept.json"));
  readOpts.add(Option("resolution", 4.0));
  readOpts.add(Option("threads", 64));

  Stage* reader(factory->createStage("readers.ept"));
  reader->addOptions(readOpts);

  StreamCallbackFilter f;

  point_count_t count(0);

  auto convertOne = [&](PointRef& point) {
    count++;
    return true;
  };

  f.setCallback(convertOne);
  f.setInput(*reader);

  Options writeOpts;
  writeOpts.add(Option("filename", "output.laz"));

  Stage* writer(factory->createStage("writers.las"));
  writer->addOptions(writeOpts);
  writer->setInput(f);

  FixedPointTable table(100);
  writer->prepare(table);
  writer->execute(table);

  cout << count << endl;
  return 0;
}

Returns 196 929 for a 112 774 395 points point cloud (and outputfile.laz has 196 929 points).
Returns 1 071 500 117 for a 17 642 072 775 points point cloud (and outputfile.laz has 1 071 500 117 points).

Using a pipeline :
pipeline.json

[
    {
        "type": "readers.ept",
        "filename": "some/url/to/ept.json",
        "resolution": 4.0
    },
    {
        "type": "writers.las",
        "filename": "outputfile.laz"
    }
]

pdal pipeline pipeline.json

outputfile.laz has 44 734 points for a 112 774 395 points point cloud.
outputfile.laz has 359 268 for a 17 642 072 775 points point cloud.

[abellgithub]

You're not reading the full resolution, so you wouldn't expect the count to be the same.

[gui2dev]

@abellgithub, my issue is that when trying not to read the full resolution using C++ API by setting the right option, it does nonetheless.

[gui2dev]

@abellgithub, the qi.m_pointCount is always the total number of point, regardless of the resolution option value.
To dig deeper, my C++ example seems to work on a small point cloud (112 million points) but not on a larger point cloud (17 billion).
Could it be a 32bits issue ? (#3121)

[abellgithub]

If this is the issue, it has already been fixed: 0425b51

…

On Fri, Jun 19, 2020 at 4:53 PM Guilhem ***@***.***> wrote: @abellgithub <https://github.com/abellgithub>, the qi.m_pointCount is always the total number of point, regardless of the resolution option value. To dig deeper, my C++ example seems to work on a small point cloud (112 million points) but not on a larger point cloud (17 billion). Could it be a 32bits issue ? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3139 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKBMMDFGPWLH7YQ7ZBZSB3RXPF5RANCNFSM4OC4QNAA> .

-- Andrew Bell andrew.bell.ia@gmail.com

[gui2dev]

@abellgithub, it's not the issue.
The issue is that the resolution option is not taken into account in my C++ example if the point cloud size is larger than 32 bits.

[abellgithub]

It's not clear that this is the problem. Your need to create a clear issue that includes ONLY the program that shows the error, the expected output and the output you believe to be incorrect.

…

On Fri, Jun 19, 2020 at 5:09 PM Guilhem ***@***.***> wrote: @abellgithub <https://github.com/abellgithub>, it's not the issue. The issue is that the resolution option is not taken into account in my C++ example if the point cloud size is larger than 32 bits. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3139 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKBMME3ZUI4NEP3AX7YGPTRXPHXLANCNFSM4OC4QNAA> .

-- Andrew Bell andrew.bell.ia@gmail.com

[gui2dev]

@abellgithub, I updated my issue.

[gui2dev]

@abellgithub, @connormanning, I found the issue.
It's related with how m_depthEnd is managed in ìnitialize()
If you do a preview() it'll go through initialize() and if you query a resolution, m_depthEnd will be incremented.
Then if you execute the same reader stage, it'll got through initialize() once again and m_depthEnd will be incremented again.