Add gdal raster as-features by dbaston · Pull Request #12970 · OSGeo/gdal

dbaston · 2025-08-27T12:50:44Z

Adds a command to convert individual raster pixels into features (points, polygons, or non-spatial). Some of this could previously be done with gdal2ogr.

apps/gdalalg_raster_as_features.cpp

coveralls · 2025-08-27T13:43:36Z

Changes unknown
when pulling ebbb0ee on dbaston:gdal-raster-as-features
into ** on OSGeo:master**.

apps/gdalalg_raster_as_features.cpp

rouault · 2025-08-27T18:08:56Z

apps/gdalalg_raster_as_features.cpp

+        while (m_row < m_window.nYSize)
+        {
+            const double *pSrcVal = static_cast<double *>(m_buf) +
+                                    (m_row * m_window.nXSize * m_bands.size() +


should make codeql happy:

Suggested change

(m_row * m_window.nXSize * m_bands.size() +

(m_bands.size() * m_row * m_window.nXSize +

apps/gdalalg_raster_as_features.cpp

gcore/gdal_priv.h

rouault · 2025-08-28T08:21:08Z

apps/gdalalg_raster_as_features.cpp

+            m_bandFields.push_back(m_defn->GetFieldIndex(fieldName));
+        }
+
+        ResetReading();


to make cppcheck happy

Suggested change

ResetReading();

GDALRasterAsFeaturesLayer::ResetReading();

jratike80 · 2025-08-28T18:15:26Z

Should we have some sort of test for pixel-is-area and pixel-is-point rasters? And should we warn users if they want to convert pixel-is-point data into polygons?

dbaston · 2025-08-29T14:34:50Z

clang static analyzer failing with

Summary of errors found:
{
  "uri": "file:///home/runner/work/gdal/gdal/gcore/gdalrasterband.cpp",
  "msg": "Forming reference to null pointer",
  "location": {
    "importance": "essential",
    "location": {
      "message": {
        "text": "Forming reference to null pointer"
      },
      "physicalLocation": {
        "artifactLocation": {
          "index": 0,
          "uri": "file:///home/runner/work/gdal/gdal/gcore/gdalrasterband.cpp"
        },
        "region": {
          "endColumn": 52,
          "endLine": 9642,
          "startColumn": 21,
          "startLine": 9642
        }
      }
    }
  }
}

Code in question:

gdal/gcore/gdalrasterband.cpp

Lines 9634 to 9644 in 4eb6fd8

    
           GDALRasterBand::WindowIteratorWrapper::WindowIteratorWrapper( 
        
               const GDALRasterBand &band) 
        
               : m_nRasterXSize(band.GetXSize()), m_nRasterYSize(band.GetYSize()), 
        
                 m_nBlockXSize(-1), m_nBlockYSize(-1) 
        
           { 
        
               // If invalid block size is reported, just use a value of 1. 
        
               CPLErrorStateBackuper state(CPLQuietErrorHandler); 
        
               band.GetBlockSize(&m_nBlockXSize, &m_nBlockYSize); 
        
               m_nBlockXSize = std::max<int>(m_nBlockXSize, 1); 
        
               m_nBlockYSize = std::max<int>(m_nBlockYSize, 1); 
        
           }

I confess I don't understand the problem.

rouault · 2025-08-29T14:42:40Z

apps/gdalalg_raster_as_features.cpp

+            m_it = m_ds.GetRasterBand(1)->IterateWindows().begin();
+            m_end = m_ds.GetRasterBand(1)->IterateWindows().end();


maybe try the following to make clang static analyzer happy ?

Suggested change

m_it = m_ds.GetRasterBand(1)->IterateWindows().begin();

m_end = m_ds.GetRasterBand(1)->IterateWindows().end();

#include <cassert>

Suggested change

m_it = m_ds.GetRasterBand(1)->IterateWindows().begin();

m_end = m_ds.GetRasterBand(1)->IterateWindows().end();

GDALRasterBand* poFirstBand = m_ds.GetRasterBand(1);

assert(poFirstBand);

m_it = poFirstBand->IterateWindows().begin();

m_end = poFirstBand->IterateWindows().end();

and this will need to be rebased on top of latest master to fix conflict

rouault · 2025-08-29T21:58:03Z

hum, CSA still no happy. maybe try the following (using plain assert(), and adding a #include <cassert>) ?

GDALRasterBand::WindowIteratorWrapper GDALRasterBand::IterateWindows() const
{
#if defined(CSA_BUILD)
    assert(this != nullptr);
#endif
    return WindowIteratorWrapper(*this);
}

otherwise we'll need to add a suppression in ci/travis/csa_common/script.sh

rouault · 2025-09-02T16:58:54Z

gcore/gdalrasterband.cpp

+    // If invalid block size is reported, just use a value of 1.
+    CPLErrorStateBackuper state(CPLQuietErrorHandler);
+#ifdef CSA_BUILD
+    assert(this);


I doubt this will be effective here as we are in a constructor, so this cannot be null. Did you consider my proposal of putting it in GDALRasterBand::IterateWindows() instead ? There this could be null if calling IterateWindows() from a null pointer

Yes, I tried that locally but it didn't seem to have an effect. This did, though I'm not confident in the diagnostic reporting from incremental builds.

ok, weird. You'll need to rebase on top of latest master, due to changes in const'ness of virtual methods

rouault · 2025-09-02T17:33:19Z

arg, I messed up something on master (now fixed) before you rebased on it. I've rebased for you and pushed on your branch, so nothing to do (but you'll need to "git fetch dbaston gdal-raster-as-features" and "git rebase dbaston/gdal-raster-as-features" if you need to apply further updates)

dbaston added the gdal_cli Anything related to the new 3.11 "gdal" CLI frontend label Aug 27, 2025

github-advanced-security bot found potential problems Aug 27, 2025

View reviewed changes

apps/gdalalg_raster_as_features.cpp Fixed Show fixed Hide fixed

apps/gdalalg_raster_as_features.cpp Fixed Show fixed Hide fixed

apps/gdalalg_raster_as_features.cpp Fixed Show fixed Hide fixed

apps/gdalalg_raster_as_features.cpp Fixed Show fixed Hide fixed

rouault reviewed Aug 27, 2025

View reviewed changes

rouault reviewed Aug 28, 2025

View reviewed changes

rouault reviewed Aug 29, 2025

View reviewed changes

dbaston force-pushed the gdal-raster-as-features branch from 4eb6fd8 to 01c34ca Compare August 29, 2025 17:56

rouault reviewed Sep 2, 2025

View reviewed changes

dbaston force-pushed the gdal-raster-as-features branch from 755b83a to 5a44cc5 Compare September 2, 2025 17:23

dbaston added 5 commits September 2, 2025 19:31

Add gdal raster as-features

8d73142

gdal raster as-features: handle degenerate rasters

99c8584

gdal raster as-features: address linter warnings

3c7a261

gdal raster as-features: appease clang scan-build

bf95675

gdal raster as-features: avoid scan-build warning

fc5cd1f

rouault force-pushed the gdal-raster-as-features branch from 5a44cc5 to fc5cd1f Compare September 2, 2025 17:32

test_gdalalg_raster_as_features.py: update test

ebbb0ee

rouault approved these changes Sep 2, 2025

View reviewed changes

rouault added this to the 3.12.0 milestone Sep 3, 2025

rouault merged commit c056f97 into OSGeo:master Sep 3, 2025
50 of 51 checks passed

	(m_row * m_window.nXSize * m_bands.size() +
	(m_bands.size() * m_row * m_window.nXSize +

		m_it = m_ds.GetRasterBand(1)->IterateWindows().begin();
		m_end = m_ds.GetRasterBand(1)->IterateWindows().end();

	m_it = m_ds.GetRasterBand(1)->IterateWindows().begin();
	m_end = m_ds.GetRasterBand(1)->IterateWindows().end();
	#include <cassert>

-            m_it = m_ds.GetRasterBand(1)->IterateWindows().begin();
-            m_end = m_ds.GetRasterBand(1)->IterateWindows().end();
+            GDALRasterBand* poFirstBand = m_ds.GetRasterBand(1);
+            assert(poFirstBand);
+            m_it = poFirstBand->IterateWindows().begin();
+            m_end = poFirstBand->IterateWindows().end();

Uh oh!

Conversation

dbaston commented Aug 27, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

coveralls commented Aug 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

rouault Aug 27, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

rouault Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

jratike80 commented Aug 28, 2025

Uh oh!

dbaston commented Aug 29, 2025

Uh oh!

rouault Aug 29, 2025

Choose a reason for hiding this comment

Uh oh!

rouault commented Aug 29, 2025

Uh oh!

rouault Sep 2, 2025

Choose a reason for hiding this comment

Uh oh!

dbaston Sep 2, 2025

Choose a reason for hiding this comment

Uh oh!

rouault Sep 2, 2025

Choose a reason for hiding this comment

Uh oh!

rouault commented Sep 2, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

coveralls commented Aug 27, 2025 •

edited

Loading