Segfaults when accessing closed datasets about node-gdal HOT 6 CLOSED

Yeah agreed. I was just talking to @brianreavis about that the other day. I was debating whether or not to fix this issue because all solutions seem messy.

It requires a null pointer check every time a destroyable object (dataset, band, ogrfeature, ogrdatasource, etc) is unwrapped. This happens in all of the object's wrapped methods and every time it is used in a method argument. I ended up going with this solution even though it's kinda nasty.

https://github.com/naturalatlas/node-gdal/blob/master/src/gdal_common.hpp#L385-L386

from node-gdal.

Whoops. That doesn't fix the whole problem. If you destroy a Dataset, all of the RasterBand pointers still are non-null, but are bogus. I'll hopefully get around to fixing this tomorrow.

Similar problem with OGR and datasources/layers:
https://github.com/naturalatlas/node-ogr/issues/2

from node-gdal.

In Rasterio, I did away with GDAL Band objects and replaced them with more featureful Numpy arrays. This also did away with dangling pointers. Would it be useful and possible to do this for Node? Maybe using https://www.npmjs.org/package/ndarray?

from node-gdal.

I don't have experience with ndarray, but it looks useful and promising. I don't think we would want to replace gdal bands with it in the binding though. If we did, we would have to read the whole raster into memory, which could be slower than desired. Also, then you would lose the ability to access overviews.

I think allowing people to easily copy a band into a ndarray would be the better option.

The OGR datasource/layer dangling pointer problem is something we can't fix by copying data, so we'll have to come up with some solution. I haven't had the time to give it much thought yet.

One option is that we could maintain a list of child Layer objects for each Datasource object so that when a datasource is destroyed we can set all the pointers to the OGRLayers to null.

from node-gdal.

Nice to see that the immediate ds.close() issue is solved:

> ds.close()
undefined
> ds.getRasterXSize()
Error: Dataset object has already been destroyed
    at repl:1:5
    at REPLServer.self.eval (repl.js:110:21)
    at Interface.<anonymous> (repl.js:239:12)
    at Interface.EventEmitter.emit (events.js:95:17)
    at Interface._onLine (readline.js:202:10)
    at Interface._line (readline.js:531:8)
    at Interface._ttyWrite (readline.js:760:14)
    at ReadStream.onkeypress (readline.js:99:10)
    at ReadStream.EventEmitter.emit (events.js:98:17)
    at emitKey (readline.js:1095:12)

from node-gdal.

Now have bands properly being disposed when datasets are closed: (cf57534, 95ed8f3)

> band.getXSize()
984
> ds.close()
undefined
> band.getXSize()
Error: RasterBand object has already been destroyed
    at repl:1:7
    at REPLServer.self.eval (repl.js:110:21)
    at Interface.<anonymous> (repl.js:239:12)
    at Interface.EventEmitter.emit (events.js:95:17)
    at Interface._onLine (readline.js:202:10)
    at Interface._line (readline.js:531:8)
    at Interface._ttyWrite (readline.js:760:14)
    at ReadStream.onkeypress (readline.js:99:10)
    at ReadStream.EventEmitter.emit (events.js:98:17)
    at emitKey (readline.js:1095:12)

from node-gdal.