Basic tools for setting up a function to be applied over a raster processing chain. The rios.applier.apply() function is the main point of entry in this module.

class rios.applier.ApplierControls[source]

Controls for the operation of rios, for use with the rios.applier.apply() function.

This object starts with default values for all controls, and has methods for setting each of them to something else.

Default values are provided for all attributes, and can then be over-ridden with the ‘set’ methods given.

Some ‘set’ methods take an optional imagename argument. If given, this shouldbe the same internal name used for a given image as in the rios.structures.FilenameAssociations objects. This is the internal name for that image, and the method will set the parameter in question for that specific image, which will over-ride the global value set when no imagename is given.

Attributes are:
  • windowxsize X size of rios block (pixels)

  • windowysize Y size of rios block (pixels)

  • overlap Number of pixels in margin for block overlaps

  • footprint rios.applier.INTERSECTION or rios.applier.UNION or rios.applier.BOUNDS_FROM_REFERENCE

  • drivername GDAL driver short name for output

  • creationoptions GDAL creation options for output

  • thematic True/False for thematic outputs

  • layernames List of layer names for outputs

  • referenceImage Image for reference projection and grid that inputs will be resampled to.

  • referencePixgrid pixelGrid for reference projection and grid

  • loggingstream file-like for logging of messages

  • progress progress object

  • statsIgnore global stats ignore value for output (i.e. null value)

  • inputnodata Over-ride of null value for input file, in reprojecting

  • calcStats True/False to signal calculate statistics and pyramids

  • omitPyramids True/False to omit pyramids when doing stats

  • overviewLevels List of level factors used when calculating output image overviews

  • overviewMinDim Minimum dimension of highest overview level

  • overviewAggType Aggregation type for calculating overviews

  • tempdir Name of directory for temp files (resampling, etc.)

  • resampleMethod String for resample method, when required (as per GDAL)

  • numThreads Deprecated. Number of parallel threads used for processing each image block

  • jobManagerType Deprecated. Which rios.parallel.jobmanager.JobManager sub-class to use for parallel processing (by name)

  • concurrency Instance of rios.structures.ConcurrencyStyle (use instead of numThreads/jobManagerType)

  • autoColorTableType Type of color table to be automatically added to thematic output rasters

  • allowOverviewsGdalwarp Allow use of overviews in input resample (dangerous, do not use)

  • approxStats Allow approx stats (much faster)

  • layerselection List of selected layer numbers for input

Options relating to vector input files
  • burnvalue Value to burn into raster from vector

  • filtersql SQL where clause used to filter vector features

  • alltouched Boolean. If True, all pixels touched are included in vector.

  • burnattribute Name of vector attribute used to supply burnvalue

  • vectorlayer Number (or name) of vector layer

  • vectordatatype Numpy datatype to use for raster created from vector

  • vectornull Rasterised vector is initialised to this value, before burning


Uses the new ConcurrencyStyle model to emulate the old JobManager concurrency. The new stuff is much better, but this allows old programs to use it without modification. Prints a deprecation warning. This routine is called automatically if the old JobManager settings have been invoked.

getOptionForImagename(option, imagename)[source]

Returns the value of a particular option for the given imagename. If only the global option has been set, then that is returned, but if a specific value has been set for the given imagename, then use that.

The imagename is the same internal name as used for the image in the rios.structures.FilenameAssociations objects.

selectInputImageLayers(layerselection, imagename=None)[source]

Set which layers are to be read from the input image(s). Default will read all layers. If imagename is given, selection will be for that image only. The layerselection parameter should be a list of layer numbers. Layer numbers follow GDAL conventions, i.e. a layer number of 1 refers to the first layer in the file. Can be much more efficient when only using a small subset of layers from the inputs.


This option is provided purely for testing purposes, and it is recommended that this never be used operationally.

In GDAL >= 2.0, the default behaviour of gdalwarp was modified so that it will use overviews during a resample to a lower resolution. By default, RIOS now switches this off again (by giving gdalwarp the ‘-ovr NONE’ switch), as this is very unreliable behaviour. Overviews can be calculated by many different methods, and the user of the file cannot tell how they were done.

In order to allow users to assess the damage done by this, we provide this option to allow resampling to use overviews. This also allows compatibility with versions of RIOS which did not switch it off, before we discovered that it was happening. To allow this, set this parameter to True, otherwise it defaults to False.

We strongly recommend against allowing gdalwarp to use overviews.

setAlltouched(alltouched, vectorname=None)[source]

Set boolean value of alltouched attribute. If alltouched is True, then pixels will count as “inside” a vector polygon if they touch the polygon, rather than only if their centre is inside. If vectorname given, then set only for that vector.

Default is False.


Set boolean value of approxStats attribute. This modifies the behaviour of calcStats by forcing it to use the pyramid layers during stats generation (much faster but only provides approximate values, not recommended for thematic rasters)

setAutoColorTableType(autoColorTableType, imagename=None)[source]

If this option is set, then thematic raster outputs will have a color table automatically generated and attached to them. The type is passed to rios.rat.genColorTable() to determine what type of automatic color table is generated.

The default type will be taken from $RIOS_DFLT_AUTOCOLORTABLETYPE if it is set. If that is not set, then the default is not to automatically attached any color table to thematic output rasters.

In practise, it is probably simpler to explicitly set the color table using the rios.rat.setColorTable() function, after creating the file, but this is an alternative.

Note that the imagename parameter can be given, in which case the autoColorTableType will only be applied to that raster.

None of this has any impact on athematic outputs.

setBurnAttribute(burnattribute, vectorname=None)[source]

Set the vector attribute name from which to get the burn value for each vector feature. If vectorname is given, set only for that vector input. Default is to use burnvalue instead of burnattribute.

setBurnValue(burnvalue, vectorname=None)[source]

Set the burn value to be used when rasterizing the input vector(s). If vectorname given, set only for that vector. Default is 1.

setCalcStats(calcStats, imagename=None)[source]

Set True to calc stats, False otherwise. If True, then statistics and pyramid layers are calculated (if supported by the driver).

Default is True.


Set the concurrency style. Argument is an instance of the rios.structures.ConcurrencyStyle class. See there for full details of how to use this.

New in version 2.0

setCreationOptions(creationoptions, imagename=None)[source]

Set a list of GDAL creation options (should match with the driver). Each list element is a string of the form “NAME=VALUE”.

Defaults are suitable for the default driver, and need to be changed if that is changed. However, if an appropriate driver-specific default environment variable ($RIOS_DFLT_CREOPT_<driver>) is given, this will be used.

setFilterSQL(filtersql, vectorname=None)[source]

Set an SQL WHERE clause which will be used to filter vector features. If vectorname is given, then set only for that vector


Set type of footprint, one of INTERSECTION, UNION or BOUNDS_FROM_REFERENCE from this module

The footprint type controls the extent of the pixel grid used for calculation within the user function, and of the output files.

Using INTERSECTION will result in the maximum extent which is wholly included in all of the input images. Using UNION results in the minimum extent which wholly includes all of the input images. If BOUNDS_FROM_REFERENCE is used, then the extent will be the same as that of the reference image or pixgrid, regardless of the extents of the various other inputs.

For both UNION and BOUNDS_FROM_REFERENCE, it is possible to have pixels which are within the extent, but outside one or more of the input files. The input data for such pixels are filled with the null value for that file. If no null value is set for that file, then zero is used.

setInputNoDataValue(nodataValue, imagename=None)[source]

Set a ‘no data’ value for input file(s). This over-rides whatever null value may be defined within the file itself, and most importantly, will over-ride when the file does not have a null value set at all.

The main reason this is important is when an input file does not have a null value, and is being reprojected on input. Since it has no null value, the resampling on the edge of any null value region in the image will risk blurring the nulls into the non-null area. Normally this is avoided if the file has a null value set. If the input file is not under the control of the user, then it cannot be set before reading, so this allows the user to over-ride it, and behave as though it had been set on the original file.

If the supplied value is a single scalar, it will apply to all bands of the input, but if it is a list of values, they will apply per-band.

If there is no reprojection on input, then this setting will have no effect on the data. However, it will be honoured by a call to info.getNoDataValueFor(), inside the user function (although that is itself discouraged).

If the imagename parameter is used, then the setting will apply only to that input, otherwise it will be applied to all inputs.


This is now deprecated. Please see setConcurrencyStyle instead.

Set which type of JobManager is to be used for parallel processing. See rios.parallel.jobmanager for details. Default is taken from $RIOS_DFLT_JOBMGRTYPE.

setLayerNames(layerNames, imagename=None)[source]

Set list of layernames to be given to the output file(s). This is not really well supported by most format drivers, and should probably be avoided. It seemed like a good idea at the time.


Set the rios logging stream to the given file-like object.

This is now deprecated (v2.0.0), and has no effect. The loggingstream is no longer used within RIOS.


This is now deprecated. Please see setConcurrencyStyle instead.

Set the number of ‘threads’ to be used when processing each block of imagery. Note that these are not threads in the technical sense, but are handled by the JobManager class, and are some form of cooperating parallel processes, depending on the type of job manager sub-class selected. See rios.parallel.jobmanager for full details. Note that this is only worth using on very computationally-intensive tasks. Default is 1, i.e. no parallel processing.

setOmitPyramids(omitPyramids, imagename=None)[source]

Set True to omit pyramid layers (i.e. overviews), False otherwise. If True, then when statistics are being calculated, pyramid layers will be omitted, otherwise they will be created at the same time.

Default is False, meaning that pyramid layers will be calculated on all output files.

setOptionForImagename(option, imagename, value)[source]

Set the given option specifically for the given imagename. This method is for internal use only. If you wish to set a particular attribute, use the corresponding ‘set’ method.

setOutputDriverName(drivername, imagename=None)[source]

Set the output driver name to the given GDAL shortname.

Note that the GDAL creation options have defaults suitable only for the default driver, so if one sets the output driver, then the creation options should be reviewed too.

In more recent versions of RIOS, the addition of driver-specific default creation options ($RIOS_DFLT_CREOPT_<driver>) allows for multiple default creation options to be set up.


Set the overlap to the given value.

Overlap is a number of pixels, and is somewhat mis-named. It refers to the amount of margin added to each block of input, so that the blocks will overlap, hence the actual amount of overlap is really more like double this value (allowing for odd and even numbers, etc).

The margin can result in pixels which are outside the extent of the given input images. These pixels will be filled with the null value for that input file, or zero if no null value is set on that file.

setOverviewAggregationType(overviewAggType, imagename=None)[source]

Set the type of aggregation used when computing overview images (i.e. pyramid layers). Normally a thematic image should be aggregated using “NEAREST”, while a continuous image should be aggregated using “AVERAGE”. When the setting is given as None, then a default is used. If using an output format which supports LAYER_TYPE, the default is based on this, but if not, it comes from the value of the environment variable $RIOS_DEFAULT_OVERVIEWAGGREGATIONTYPE.

This method should usually be used to set when writing an output to a format which does not support LAYER_TYPE, and which is not appropriate for the setting given by the environment default.

setOverviewLevels(overviewLevels, imagename=None)[source]

Set the overview levels to be used on output images (i.e. pyramid layers). Levels are specified as a list of integer factors, with the same meanings as given to the gdaladdo command.

setOverviewMinDim(overviewMinDim, imagename=None)[source]

Set minimum dimension allowed for output overview. Overview levels (i.e. pyramid layers) will be calculated as per the overviewLevels list of factors, but only until the minimum dimension falls below the value of overviewMinDim


Set the progress display object. Default is no progress object.

The progress object should be an instance of one of the classes from rios.cuiprogress, and is used to generate a simple progress indicator showing the percentage completed.


Set the name of the image to use for the reference pixel grid and projection. If neither referenceImage nor referencePixgrid are set, then no resampling will be allowed. Only set one of referenceImage or referencePixgrid.

The reference image can be given as either the internal name, as given on infiles, or the external filename. The internal name is preferred, and consistent with other usage, but the filename is allowed for backward compatibility.


Set the reference pixel grid. If neither referenceImage nor referencePixgrid are set, then no resampling will be allowed. Only set one of referenceImage or referencePixgrid. The referencePixgrid argument is of type rios.pixelgrid.PixelGridDefn.

setResampleMethod(resampleMethod, imagename=None)[source]

Set resample method to be used for resampling of input files. Possible options are those defined by gdalwarp, i.e. ‘near’, ‘bilinear’, ‘cubic’, ‘cubicspline’, ‘lanczos’.

setStatsIgnore(statsIgnore, imagename=None)[source]

Set the ‘no data’ value for the output files (also known as the ‘null’ value). This value will be written into the output files, and will thus be ignored when calculating statistics, histograms and overviews (pyramid layers) on those files.

If this value is given as None, then no null value will be set on output files.

If imagename is given, the setting will only apply to that image, otherwise it will apply to all output files.

There is currently no mechanism to set different null values for different layers in an output file.

The default value is 0. This was probably a bad idea, but to avoid breaking old scripts, we are not likely to change this.


Set directory to use for temporary files for resampling, etc.

Default is ‘.’ (i.e. current directory).

setThematic(thematicFlag, imagename=None)[source]

Boolean flag to indicate whether the output file is thematic. A value of True means the output will be set as thematic, although this may not be supported by the output format driver.

Default is False (i.e. not thematic).

setVectorDatatype(vectordatatype, vectorname=None)[source]

Set numpy datatype to use for rasterized vectors. If vectorname given, set only for that vector.

Default is numpy.uint8

setVectorNull(vectornull, vectorname=None)[source]

Set the vector null value. This is used to initialise the rasterised vector, before burning in the burn value. This is of most importance when burning values from a vector attribute column, as this should be a distinct value from any of the values in the column. If this is not so, then polygons can end up blending with the background, resulting in incorrect answers.

Default is 0

setVectorlayer(vectorlayer, vectorname=None)[source]

Set number/name of vector layer, for vector formats which have multiple layers. Not required for plain shapefiles. Can be either a layer number (start at zero) or a layer name. If vectorname given, set only for that vector.

setWindowSize(windowxsize, windowysize)[source]

Sets the X and Y size of the blocks used in one call. Images are processed in blocks (windows) of ‘windowxsize’ columns, and ‘windowysize’ rows.


Set the X size of the blocks used. Images are processed in blocks (windows) of ‘windowxsize’ columns, and ‘windowysize’ rows.


Set the Y size of the blocks used. Images are processed in blocks (windows) of ‘windowxsize’ columns, and ‘windowysize’ rows.

class rios.applier.ApplierProgress(controls, numBlocks)[source]

Wrapper around the controls progress object, just to simplify the update call, and keeping track of whether the percentage has changed.

rios.applier.apply(userFunction, infiles, outfiles, otherArgs=None, controls=None)[source]

Apply the given ‘userFunction’ to the given input and output files.

infiles and outfiles are rios.structures.FilenameAssociations objects to define associations between internal variable names and external filenames, for the raster file inputs and outputs.

otherArgs is an object of extra arguments to be passed to the userFunction, each with a sensible name on the object. These can be either input or output arguments, entirely at the discretion of userFunction(). otherArgs should be in instance of rios.structures.OtherInputs

The userFunction has the following call sequence:

userFunction(info, inputs, outputs)


userFunction(info, inputs, outputs, otherArgs)

if otherArgs is not None.

inputs and outputs are objects in which there are named attributes with the same names as those given in the infiles and outfiles objects. In the inputs and outputs objects, available inside userFunction, these attributes contain numpy arrays of data read from/written to the corresponding image file.

If the attributes given in the infiles or outfiles objects are lists of filenames, the the corresponding attributes of the inputs and outputs objects inside the applied function will be lists of image data blocks instead of single blocks.

The numpy arrays are always 3-d arrays, with shape:

(numBands, numRows, numCols)

The datatype of the output image(s) is determined directly from the datatype of the numpy arrays in the outputs object.

The info object contains many useful details about the processing, and will always be passed to the userFunction. It can, of course, be ignored. It is an instance of the rios.readerinfo.ReaderInfo class.

The controls argument, if given, is an instance of the rios.applier.ApplierControls class, which allows control of various aspects of the reading and writing of images. See the class documentation for further details.

The apply function returns a rios.structures.ApplierReturn object (new in version 2.0).

There is a page dedicated to Applier Examples.

rios.applier.apply_multipleCompute(userFunction, infiles, outfiles, otherArgs, controls, allInfo, workinggrid, blockList)[source]

Called internally from the apply() function. Not to be called directly.

Apply function for the multiple compute cases. Starts a number of compute workers, each of which calls the user function on inputs and creates outputs, which this function writes to the output files.

rios.applier.apply_singleCompute(userFunction, infiles, outfiles, otherArgs, controls, allInfo, workinggrid, blockList, outBlockBuffer, inBlockBuffer, workerID, forceExit)[source]

Called internally from the apply() function. Not to be called directly.

Apply function for simplest configuration, with no compute concurrency. Does have possible read concurrency.

This function is also called for each compute worker in the batch-oriented compute worker styles, where each worker is an instance of a single-compute case.

rios.applier.checkAllMatch(pixgridList, refPixGrid)[source]

Returns whether any resampling necessary to match reference dataset.

Use as a check if no resampling is done that we can proceed ok.

rios.applier.makeBlockList(workinggrid, controls)[source]

Divide the working grid area into blocks. Return a list of ApplierBlockDefn objects

rios.applier.makeWorkingGrid(infiles, allInfo, controls)[source]

Work out the projection and extent of the working grid.

Return a PixelGridDefn object representing it.


Open all input files and create an ImageInfo (or VectorFileInfo) object for each. Return a dictionary of them, keyed by their position within infiles, i.e. (symbolicName, SeqNum).


Report the given WorkerExceptionRecord object to stderr