 NAME
 DESCRIPTION
 SYNOPSIS
 VECTORIZATION/THREADING: METHOD AND NOMENCLATURE
 VARIABLES
 FUNCTIONS
 barf
 pdl
 null
 nullcreate
 nelem
 dims
 shape
 ndims
 getndims
 dim
 getdim
 topdl
 PDL::get_datatype
 howbig
 get_dataref
 upd_data
 PDL::threadids
 doflow
 flows
 PDL::new
 copy
 PDL::hdr_copy
 PDL::unwind
 PDL::make_physical
 dummy
 clump
 thread_define
 PDL::thread
 diagonal
 PDL::thread1
 PDL::thread2
 PDL::thread3
 sever
 PDL::info
 approx
 mslice
 nslice_if_pdl
 nslice
 inplace
 is_inplace
 set_inplace
 new_or_inplace
 PDL::new_from_specification
 isnull
 isempty
 zeroes
 zeros
 ones
 reshape
 squeeze
 flat
 convert
 Datatype_conversions
 byte
 short
 ushort
 long
 longlong
 float
 double
 type
 list
 unpdl
 listindices
 set
 at
 sclr
 cat
 dog
 gethdr
 hdr
 fhdr
 sethdr
 hdrcpy
 hcpy
 set_autopthread_targ
 get_autopthread_targ
 get_autopthread_actual
 set_autopthread_size
 get_autopthread_size
 AUTHOR
NAME
PDL::Core  fundamental PDL functionality and vectorization/threading
DESCRIPTION
Methods and functions for type conversions, PDL creation, type conversion, threading etc.
SYNOPSIS
use PDL::Core; # Normal routines
use PDL::Core ':Internal'; # Hairy routines
VECTORIZATION/THREADING: METHOD AND NOMENCLATURE
PDL provides vectorized operations via a builtin engine. Vectorization is called "threading" for historical reasons. The threading engine implements simple rules for each operation.
Each PDL object has a "shape" that is a generalized Ndimensional rectangle defined by a "dim list" of sizes in an arbitrary set of dimensions. A PDL with shape 2x3 has 6 elements and is said to be twodimensional, or may be referred to as a 2x3PDL. The dimensions are indexed numerically starting at 0, so a 2x3PDL has a dimension 0 (or "dim 0") with size 2 and a 1 dimension (or "dim 1") with size 3.
PDL generalizes *all* mathematical operations with the notion of "active dims": each operator has zero or more active dims that are used in carrying out the operation. Simple scalar operations like scalar multiplication ('*') have 0 active dims. More complicated operators can have more active dims. For example, matrix multiplication ('x') has 2 active dims. Additional dims are automatically vectorized across  e.g. multiplying a 2x5PDL with a 2x5PDL requires 10 simple multiplication operations, and yields a 2x5PDL result.
Threading rules
In any PDL expression, the active dims appropriate for each operator are used starting at the 0 dim and working forward through the dim list of each object. All additional dims after the active dims are "thread dims". The thread dims do not have to agree exactly: they are coerced to agree according to simple rules:
Null PDLs match any dim list (see below).
Dims with sizes other than 1 must all agree in size.
Dims of size 1 are expanded as necessary.
Missing dims are expanded appropriately.
The "size 1" rule implements "generalized scalar" operation, by analogy to scalar multiplication. The "missing dims" rule acknowledges the ambiguity between a missing dim and a dim of size 1.
Null PDLs
PDLs on the lefthand side of assignment can have the special value "Null". A null PDL has no dim list and no set size; its shape is determined by the computed shape of the expression being assigned to it. Null PDLs contain no values and can only be assigned to. When assigned to (e.g. via the .=
operator), they cease to be null PDLs.
To create a null PDL, use PDL>null()
.
Empty PDLs
PDLs can represent the empty set using "structured Empty" variables. An empty PDL is not a null PDL.
Any dim of a PDL can be set explicitly to size 0. If so, the PDL contains zero values (because the total number of values is the product of all the sizes in the PDL's shape or dimlist).
Scalar PDLs are zerodimensional and have no entries in the dim list, so they cannot be empty. 1D and higher PDLs can be empty. Empty PDLs are useful for set operations, and are most commonly encountered in the output from selection operators such as which and whichND. Not all empty PDLs have the same threading properties  e.g. a 2x0PDL represents a collection of 2vectors that happens to contain no elements, while a simple 0PDL represents a collection of scalar values (that also happens to contain no elements).
Note that 0 dims are not adjustable via the threading rules  a dim with size 0 can only match a corresponding dim of size 0 or 1.
Thread rules and assignments
Versions of PDL through 2.4.10 have some irregularity with threading and assignments. Currently the threading engine performs a full expansion of both sides of the computed assignment operator .=
(which assigns values to a preexisting PDL). This leads to counterintuitive behavior in some cases:
Generalized scalars and computed assignment
If the PDL on the lefthand side of
.=
has a dim of size 1, it can be treated as a generalized scalar, as in:$a = sequence(2,3); $b = zeroes(1,3); $b .= $a;
In this case,
$b
is automatically treated as a 2x3PDL during the threading operation, but half of the values from$a
silently disappear. The output is, as Kernighan and Ritchie would say, "undefined".Further, if the value on the right of
.=
is empty, then.=
becomes a silent noop:$a = zeroes(0); $b = zeroes(1); $b .= $a+1; print $b;
will print
[0]
. In this case, "$a+1" is empty, and "$b" is a generalized scalar that is adjusted to be empty, so the assignment is carried out for zero elements (a noop).Both of these behaviors are considered harmful and should not be relied upon: they may be patched away in a future version of PDL.
Empty PDLs and generalized scalars
Generalized scalars (PDLs with a dim of size 1) can match any size in the corresponding dim, including 0. Thus,
$a = ones(2,0); $b = sequence(2,1); $c = $a * $b; print $c;
prints
Empty[2,0]
.This behavior is counterintuitive but desirable, and will be preserved in future versions of PDL.
VARIABLES
These are important variables of global scope and are placed in the PDL namespace.
$PDL::debug
When true, PDL debugging information is printed.
$PDL::verbose
When true, PDL functions provide chatty information.
$PDL::use_commas
Whether to insert commas when printing pdls
$PDL::floatformat
, $PDL::doubleformat
The default print format for floats and doubles, repectively. The default default values are:
$PDL::floatformat = "%7g";
$PDL::doubleformat = "%10.8g";
$PDL::undefval
The value to use instead of undef
when creating pdls.
$PDL::toolongtoprint
The maximal size pdls to print (defaults to 10000 elements)
FUNCTIONS
barf
Standard error reporting routine for PDL.
barf()
is the routine PDL modules should call to report errors. This is because barf()
will report the error as coming from the correct line in the module user's script rather than in the PDL module.
For now, barf just calls Carp::confess()
Remember barf()
is your friend. *Use* it!
At the perl level:
barf("User has too low an IQ!");
In C or XS code:
barf("You have made %d errors", count);
Note: this is one of the few functions ALWAYS exported by PDL::Core
pdl
PDL constructor  creates new piddle from perl scalars/arrays, piddles, and strings
$a = pdl(SCALARARRAY REFERENCEARRAYSTRING);
$a = pdl [1..10]; # 1D array
$a = pdl ([1..10]); # 1D array
$a = pdl (1,2,3,4); # Ditto
$b = pdl [[1,2,3],[4,5,6]]; # 2D 3x2 array
$b = pdl "[[1,2,3],[4,5,6]]"; # Ditto (slower)
$b = pdl "[1 2 3; 4 5 6]"; # Ditto
$b = pdl q[1 2 3; 4 5 6]; # Ditto, using the q quote operator
$b = pdl "1 2 3; 4 5 6"; # Ditto, less obvious, but still works
$b = pdl 42 # 0dimensional scalar
$c = pdl $a; # Make a new copy
$a = pdl([1,2,3],[4,5,6]); # 2D
$a = pdl([[1,2,3],[4,5,6]]); # 2D
Note the last two are equivalent  a list is automatically converted to a list reference for syntactic convenience. i.e. you can omit the outer []
You can mix and match arrays, array refs, and PDLs in your argument list, and pdl
will sort them out. You get back a PDL whose last (slowest running) dim runs across the top level of the list you hand in, and whose first (fastest running) dim runs across the deepest level that you supply. At the moment, you cannot mix and match those arguments with string arguments, though we can't imagine a situation in which you would really want to do that.
The string version of pdl also allows you to use the strings bad
, inf
, and nan
, and it will insert the values that you mean (and set the bad flag if you use bad
). You can mix and match case, though you shouldn't. Here are some examples:
$bad = pdl q[1 2 3 bad 5 6]; # Set fourth element to the bad value
$bad = pdl q[1 2 3 BAD 5 6]; # ditto
$bad = pdl q[1 2 inf bad 5]; # now third element is IEEE infinite value
$bad = pdl q[nan 2 inf inf]; # first value is IEEE nan value
The default constructor uses IEEE doubleprecision floating point numbers. You can use other types, but you will get a warning if you try to use nan
with integer types (it will be replaced with the bad
value) and you will get a fatal error if you try to use inf
.
Throwing a PDL into the mix has the same effect as throwing in a list ref:
pdl(pdl(1,2),[3,4])
is the same as
pdl([1,2],[3,4]).
All of the dimensions in the list are "paddedout" with undefval to meet the widest dim in the list, so (e.g.)
$a = pdl([[1,2,3],[2]])
gives you the same answer as
$a = pdl([[1,2,3],[2,undef,undef]]);
pdl()
is a functional synonym for the 'new' constructor, e.g.:
$x = new PDL [1..10];
In order to control how undefs are handled in converting from perl lists to PDLs, one can set the variable $PDL::undefval
. For example:
$foo = [[1,2,undef],[undef,3,4]];
$PDL::undefval = 999;
$f = pdl $foo;
print $f
[
[ 1 2 999]
[999 3 4]
]
$PDL::undefval
defaults to zero.
As a final note, if you include an Empty PDL in the list of objects to construct into a PDL, it is kept as a placeholder pane  so if you feed in (say) 7 objects, you get a size of 7 in the 0th dim of the output PDL. The placeholder panes are completely padded out. But if you feed in only a single Empty PDL, you get back the Empty PDL (no padding).
null
Returns a 'null' piddle.
$x = null;
null()
has a special meaning to PDL::PP. It is used to flag a special kind of empty piddle, which can grow to appropriate dimensions to store a result (as opposed to storing a result in an existing piddle).
pdl> sumover sequence(10,10), $ans=null;p $ans
[45 145 245 345 445 545 645 745 845 945]
nullcreate
Returns a 'null' piddle.
$x = PDL>nullcreate($arg)
This is an routine used by many of the threading primitives (i.e. sumover, minimum, etc.) to generate a null piddle for the function's output that will behave properly for derived (or subclassed) PDL objects.
For the above usage: If $arg
is a PDL, or a derived PDL, then $arg>null
is returned. If $arg
is a scalar (i.e. a zerodimensional PDL) then PDL>null
is returned.
PDL::Derived>nullcreate(10)
returns PDL::Derived>null.
PDL>nullcreate($pdlderived)
returns $pdlderived>null.
nelem
Return the number of elements in a piddle
$n = nelem($piddle); $n = $piddle>nelem;
$mean = sum($data)/nelem($data);
dims
Return piddle dimensions as a perl list
@dims = $piddle>dims; @dims = dims($piddle);
pdl> p @tmp = dims zeroes 10,3,22
10 3 22
See also shape which returns a piddle instead.
shape
Return piddle dimensions as a piddle
$shape = $piddle>shape; $shape = shape($piddle);
pdl> p $shape = shape zeroes 10,3,22
[10 3 22]
See also dims which returns a perl list.
ndims
Returns the number of dimensions in a piddle. Alias for getndims.
getndims
Returns the number of dimensions in a piddle
$ndims = $piddle>getndims;
pdl> p zeroes(10,3,22)>getndims
3
dim
Returns the size of the given dimension of a piddle. Alias for getdim.
getdim
Returns the size of the given dimension.
$dim0 = $piddle>getdim(0);
pdl> p zeroes(10,3,22)>getdim(1)
3
Negative indices count from the end of the dims array. Indices beyond the end will return a size of 1. This reflects the idea that any pdl is equivalent to an infinitely dimensional array in which only a finite number of dimensions have a size different from one. For example, in that sense a 3D piddle of shape [3,5,2] is equivalent to a [3,5,2,1,1,1,1,1,....] piddle. Accordingly,
print $a>getdim(10000);
will print 1 for most practically encountered piddles.
topdl
alternate piddle constructor  ensures arg is a piddle
$a = topdl(SCALARARRAY REFERENCEARRAY);
The difference between pdl() and topdl()
is that the latter will just 'fall through' if the argument is already a piddle. It will return a reference and NOT a new copy.
This is particulary useful if you are writing a function which is doing some fiddling with internals and assumes a piddle argument (e.g. for method calls). Using topdl()
will ensure nothing breaks if passed with '2'.
Note that topdl()
is not exported by default (see example below for usage).
use PDL::Core ':Internal'; # use the internal routines of
# the Core module
$a = topdl 43; # $a is piddle with value '43'
$b = topdl $piddle; # fall through
$a = topdl (1,2,3,4); # Convert 1D array
PDL::get_datatype
Internal: Return the numeric value identifying the piddle datatype
$x = $piddle>get_datatype;
Mainly used for internal routines.
NOTE: get_datatype returns 'just a number' not any special type object, unlike type.
howbig
Returns the sizeof a piddle datatype in bytes.
Note that howbig()
is not exported by default (see example below for usage).
use PDL::Core ':Internal'; # use the internal routines of
# the Core module
$size = howbig($piddle>get_datatype);
Mainly used for internal routines.
NOTE: NOT a method! This is because get_datatype returns 'just a number' not any special object.
pdl> p howbig(ushort([1..10])>get_datatype)
2
get_dataref
Return the internal data for a piddle, as a perl SCALAR ref.
Most piddles hold their internal data in a packed perl string, to take advantage of perl's memory management. This gives you direct access to the string, which is handy when you need to manipulate the binary data directly (e.g. for file I/O). If you modify the string, you'll need to call upd_data afterward, to make sure that the piddle points to the new location of the underlying perl variable.
Calling get_dataref
automatically physicalizes your piddle (see make_physical). You definitely don't want to do anything to the SV to truncate or deallocate the string, unless you correspondingly call reshape to make the PDL match its new data dimension.
You definitely don't want to use get_dataref unless you know what you are doing (or are trying to find out): you can end up scrozzling memory if you shrink or eliminate the string representation of the variable. Here be dragons.
upd_data
Update the data pointer in a piddle to match its perl SV.
This is useful if you've been monkeying with the packed string representation of the PDL, which you probably shouldn't be doing anyway. (see get_dataref.)
PDL::threadids
Returns the piddle thread IDs as a perl list
Note that threadids()
is not exported by default (see example below for usage).
use PDL::Core ':Internal'; # use the internal routines of
# the Core module
@ids = threadids $piddle;
doflow
Turn on/off dataflow
$x>doflow; doflow($x);
flows
Whether or not a piddle is indulging in dataflow
something if $x>flows; $hmm = flows($x);
PDL::new
new piddle constructor method
$x = PDL>new(SCALARARRAYARRAY REFSTRING);
$x = PDL>new(42); # new from a Perl scalar
$x = new PDL 42; # ditto
$y = PDL>new(@list_of_vals); # new from Perl list
$y = new PDL @list_of_vals; # ditto
$z = PDL>new(\@list_of_vals); # new from Perl list reference
$w = PDL>new("[1 2 3]"); # new from Perl string, using
# Matlab constructor syntax
Constructs piddle from perl numbers and lists and strings with Matlab/Octave style constructor syntax.
The string input is fairly versatile though not performance optimized. The goal is to make it easy to copy and paste code from PDL output and to offer a familiar Matlab syntax for piddle construction. As of May, 2010, it is a new feature, so feel free to report bugs or suggest new features. See documentation for pdl for more examples of usage.
copy
Make a physical copy of a piddle
$new = $old>copy;
Since $new = $old
just makes a new reference, the copy
method is provided to allow real independent copies to be made.
PDL::hdr_copy
Return an explicit copy of the header of a PDL.
hdr_copy is just a wrapper for the internal routine _hdr_copy, which takes the hash ref itself. That is the routine which is used to make copies of the header during normal operations if the hdrcpy() flag of a PDL is set.
Generalpurpose deep copies are expensive in perl, so some simple optimization happens:
If the header is a tied array or a blessed hash ref with an associated method called copy
, then that >copy method is called. Otherwise, all elements of the hash are explicitly copied. References are recursively deep copied.
This routine seems to leak memory.
PDL::unwind
Return a piddle which is the same as the argument except that all threadids have been removed.
$y = $x>unwind;
PDL::make_physical
Make sure the data portion of a piddle can be accessed from XS code.
$a>make_physical;
$a>call_my_xs_method;
Ensures that a piddle gets its own allocated copy of data. This obviously implies that there are certain piddles which do not have their own data. These are so called virtual piddles that make use of the vaffine optimisation (see PDL::Indexing). They do not have their own copy of data but instead store only access information to some (or all) of another piddle's data.
Note: this function should not be used unless absolutely neccessary since otherwise memory requirements might be severly increased. Instead of writing your own XS code with the need to call make_physical
you might want to consider using the PDL preprocessor (see PDL::PP) which can be used to transparently access virtual piddles without the need to physicalise them (though there are exceptions).
dummy
Insert a 'dummy dimension' of given length (defaults to 1)
No relation to the 'Dungeon Dimensions' in Discworld!
Negative positions specify relative to last dimension, i.e. dummy(1)
appends one dimension at end, dummy(2)
inserts a dummy dimension in front of the last dim, etc.
If you specify a dimension position larger than the existing dimension list of your PDL, the PDL gets automagically padded with extra dummy dimensions so that you get the dim you asked for, in the slot you asked for. This could cause you trouble if, for example, you ask for $a>dummy(5000,1) because $a will get 5,000 dimensions, each of rank 1.
Because padding at the beginning of the dimension list moves existing dimensions from slot to slot, it's considered unsafe, so automagic padding doesn't work for large negative indices  only for large positive indices.
$y = $x>dummy($position[,$dimsize]);
pdl> p sequence(3)>dummy(0,3)
[
[0 0 0]
[1 1 1]
[2 2 2]
]
pdl> p sequence(3)>dummy(3,2)
[
[
[0 1 2]
]
[
[0 1 2]
]
]
pdl> p sequence(3)>dummy(3,2)
Runtime error: PDL: For safety, <pos> < (dims+1) forbidden in dummy. min=2, pos=3
clump
"clumps" several dimensions into one large dimension
If called with one argument $n
clumps the first $n
dimensions into one. For example, if $a
has dimensions (5,3,4)
then after
$b = $a>clump(2); # Clump 2 first dimensions
the variable $b
will have dimensions (15,4)
and the element $b>at(7,3)
refers to the element $a>at(1,2,3)
.
Use clump(1)
to flatten a piddle. The method flat is provided as a convenient alias.
Clumping with a negative dimension in general leaves that many dimensions behind  e.g. clump(2) clumps all of the first few dimensions into a single one, leaving a 2D piddle.
If clump
is called with an index list with more than one element it is treated as a list of dimensions that should be clumped together into one. The resulting clumped dim is placed at the position of the lowest index in the list. This convention ensures that clump
does the expected thing in the usual cases. The following example demonstrates typical usage:
$a = sequence 2,3,3,3,5; # 5D piddle
$c = $a>clump(1..3); # clump all the dims 1 to 3 into one
print $c>info; # resulting 3D piddle has clumped dim at pos 1
PDL: Double D [2,27,5]
thread_define
define functions that support threading at the perl level
thread_define 'tline(a(n);b(n))', over {
line $_[0], $_[1]; # make line compliant with threading
};
thread_define
provides some support for threading (see PDL::Indexing) at the perl level. It allows you to do things for which you normally would have resorted to PDL::PP (see PDL::PP); however, it is most useful to wrap existing perl functions so that the new routine supports PDL threading.
thread_define
is used to define new threading aware functions. Its first argument is a symbolic repesentation of the new function to be defined. The string is composed of the name of the new function followed by its signature (see PDL::Indexing and PDL::PP) in parentheses. The second argument is a subroutine that will be called with the slices of the actual runtime arguments as specified by its signature. Correct dimension sizes and minimal number of dimensions for all arguments will be checked (assuming the rules of PDL threading, see PDL::Indexing).
The actual work is done by the signature
class which parses the signature string, does runtime dimension checks and the routine threadover
that generates the loop over all appropriate slices of pdl arguments and creates pdls as needed.
Similar to pp_def
and its OtherPars
option it is possible to define the new function so that it accepts normal perl args as well as piddles. You do this by using the NOtherPars
parameter in the signature. The number of NOtherPars
specified will be passed unaltered into the subroutine given as the second argument of thread_define
. Let's illustrate this with an example:
PDL::thread_define 'triangles(inda();indb();indc()), NOtherPars => 2',
PDL::over {
${$_[3]} .= $_[4].join(',',map {$_>at} @_[0..2]).",1,\n";
};
This defines a function triangles
that takes 3 piddles as input plus 2 arguments which are passed into the routine unaltered. This routine is used to collect lists of indices into a perl scalar that is passed by reference. Each line is preceded by a prefix passed as $_[4]
. Here is typical usage:
$txt = '';
triangles(pdl(1,2,3),pdl(1),pdl(0),\$txt," "x10);
print $txt;
resulting in the following output
1,1,0,1,
2,1,0,1,
3,1,0,1,
which is used in PDL::Graphics::TriD::VRML to generate VRML output.
Currently, this is probably not much more than a POP (proof of principle) but is hoped to be useful enough for some real life work.
Check PDL::PP for the format of the signature. Currently, the [t]
qualifier and all type qualifiers are ignored.
PDL::thread
Use explicit threading over specified dimensions (see also PDL::Indexing)
$b = $a>thread($dim,[$dim1,...])
$a = zeroes 3,4,5;
$b = $a>thread(2,0);
Same as PDL::thread1, i.e. uses thread id 1.
diagonal
Returns the multidimensional diagonal over the specified dimensions.
$d = $x>diagonal(dim1, dim2,...)
pdl> $a = zeroes(3,3,3);
pdl> ($b = $a>diagonal(0,1))++;
pdl> p $a
[
[
[1 0 0]
[0 1 0]
[0 0 1]
]
[
[1 0 0]
[0 1 0]
[0 0 1]
]
[
[1 0 0]
[0 1 0]
[0 0 1]
]
]
PDL::thread1
Explicit threading over specified dims using thread id 1.
$xx = $x>thread1(3,1)
Wibble
Convenience function interfacing to PDL::Slices::threadI.
PDL::thread2
Explicit threading over specified dims using thread id 2.
$xx = $x>thread2(3,1)
Wibble
Convenience function interfacing to PDL::Slices::threadI.
PDL::thread3
Explicit threading over specified dims using thread id 3.
$xx = $x>thread3(3,1)
Wibble
Convenience function interfacing to PDL::Slices::threadI.
sever
sever any links of this piddle to parent piddles
In PDL it is possible for a piddle to be just another view into another piddle's data. In that case we call this piddle a virtual piddle and the original piddle owning the data its parent. In other languages these alternate views sometimes run by names such as alias or smart reference.
Typical functions that return such piddles are slice
, xchg
, index
, etc. Sometimes, however, you would like to separate the virtual piddle from its parent's data and just give it a life of its own (so that manipulation of its data doesn't change the parent). This is simply achieved by using sever
. For example,
$a = $pdl>index(pdl(0,3,7))>sever;
$a++; # important: $pdl is not modified!
In many (but not all) circumstances it acts therefore similar to copy. However, in general performance is better with sever
and secondly, sever
doesn't lead to futile copying when used on piddles that already have their own data. On the other hand, if you really want to make sure to work on a copy of a piddle use copy.
$a = zeroes(20);
$a>sever; # NOOP since $a is already its own boss!
Again note: sever
is not the same as copy! For example,
$a = zeroes(1); # $a does not have a parent, i.e. it is not a slice etc
$b = $a>sever; # $b is now pointing to the same piddle as $a
$b++;
print $a;
[1]
but
$a = zeroes(1);
$b = $a>copy; # $b is now pointing to a new piddle
$b++;
print $a;
[0]
PDL::info
Return formatted information about a piddle.
$x>info($format_string);
print $x>info("Type: %T Dim: %15D State: %S");
Returns a string with info about a piddle. Takes an optional argument to specify the format of information a la sprintf. Format specifiers are in the form %<width><letter>
where the width is optional and the letter is one of
 T

Type
 D

Formatted Dimensions
 F

Dataflow status
 S

Some internal flags (P=physical,V=Vaffine,C=changed,B=may contain bad data)
 C

Class of this piddle, i.e.
ref $pdl
 A

Address of the piddle struct as a unique identifier
 M

Calculated memory consumption of this piddle's data area
approx
test for approximately equal values (relaxed ==
)
# ok if all corresponding values in
# piddles are within 1e8 of each other
print "ok\n" if all approx $a, $b, 1e8;
approx
is a relaxed form of the ==
operator and often more appropriate for floating point types (float
and double
).
Usage:
$res = approx $a, $b [, $eps]
The optional parameter $eps
is remembered across invocations and initially set to 1e6, e.g.
approx $a, $b; # last $eps used (1e6 initially)
approx $a, $b, 1e10; # 1e10
approx $a, $b; # also 1e10
mslice
Convenience interface to slice, allowing easier inclusion of dimensions in perl code.
$a = $x>mslice(...);
# below is the same as $x>slice("5:7,:,3:4:2")
$a = $x>mslice([5,7],X,[3,4,2]);
nslice_if_pdl
If $self
is a PDL, then calls nslice
with all but the last argument, otherwise $self>($_[1]) is called where $_[1} is the original argument string found during PDL::NiceSlice filtering.
$a = $x>nslice_if_pdl(...,'(args)');
nslice
Internally used interface to slice and dice that is the runtime part of the PDL::NiceSlice implementation.
$a = $x>nslice(...);
# below is the same as $x>slice("5:7,:,3:4:2")
$a = $x>nslice([5,7],X,[3,4,2]);
It implements a superset of mslice's features. Should probably not be used in your scripts. Rather resort to the PDL::NiceSlice interface.
inplace
Flag a piddle so that the next operation is done 'in place'
somefunc($x>inplace); somefunc(inplace $x);
In most cases one likes to use the syntax $y = f($x)
, however in many case the operation f()
can be done correctly 'in place', i.e. without making a new copy of the data for output. To make it easy to use this, we write f()
in such a way that it operates inplace, and use inplace
to hint that a new copy should be disabled. This also makes for clear syntax.
Obviously this will not work for all functions, and if in doubt see the function's documentation. However one can assume this is true for all elemental functions (i.e. those which just operate array element by array element like log10
).
pdl> $x = xvals zeroes 10;
pdl> log10(inplace $x)
pdl> p $x
[inf 0 0.30103 0.47712125 0.60205999 0.69897 0.77815125 0.84509804 0.90308999 0.95424251]
is_inplace
Test the inplace flag on a piddle
$out = ($in>is_inplace) ? $in : zeroes($in);
$in>set_inplace(0)
Provides access to the inplace hint flag, within the perl millieu. That way functions you write can be inplace aware... If given an argument the inplace flag will be set or unset depending on the value at the same time. Can be used for shortcut tests that delete the inplace flag while testing:
$out = ($in>is_inplace(0)) ? $in : zeroes($in); # test & unset!
set_inplace
Set the inplace flag on a piddle
$out = ($in>is_inplace) ? $in : zeroes($in);
$in>set_inplace(0);
Provides access to the inplace hint flag, within the perl millieu. Useful mainly for turning it OFF, as inplace turns it ON more conveniently.
new_or_inplace
$a = new_or_inplace(shift());
$a = new_or_inplace(shift(),$preferred_type);
Return back either the argument pdl or a copy of it depending on whether it be flagged inplace or no. Handy for building inplaceaware functions.
If you specify a preferred type (must be one of the usual PDL type strings, a list ref containing several of them, or a string containing several of them), then the copy is coerced into the first preferred type listed if it is not already one of the preferred types.
Note that if the inplace flag is set, no coersion happens even if you specify a preferred type.
PDL::new_from_specification
Internal method: create piddle by specification
This is the argument processing method called by zeroes and some other functions which constructs piddles from argument lists of the form:
[type], $nx, $ny, $nz,...
For $nx
, $ny
, etc. 0 and 1D piddles are allowed. Giving those has the same effect as if saying $arg>list
, e.g.
1, pdl(5,2), 4
is equivalent to
1, 5, 2, 4
Note, however, that in all functions using new_from_specification
calling func $piddle
will probably not do what you want. So to play safe use (e.g. with zeroes)
$pdl = zeroes $dimpdl>list;
Calling
$pdl = zeroes $dimpdl;
will rather be equivalent to
$pdl = zeroes $dimpdl>dims;
However,
$pdl = zeroes ushort, $dimpdl;
will again do what you intended since it is interpreted as if you had said
$pdl = zeroes ushort, $dimpdl>list;
This is unfortunate and confusing but no good solution seems obvious that would not break existing scripts.
isnull
Test whether a piddle is null
croak("Input piddle mustn't be null!")
if $input_piddle>isnull;
This function returns 1 if the piddle is null, zero if it is not. The purpose of null piddles is to "tell" any PDL::PP methods to allocate new memory for an output piddle, but only when that PDL::PP method is called in fullarg form. Of course, there's no reason you couldn't commandeer the special value for your own purposes, for which this test function would prove most helpful. But in general, you shouldn't need to test for a piddle's nullness.
See "Null PDLs" for more information.
isempty
Test whether a piddle is empty
print "The piddle has zero dimension\n" if $pdl>isempty;
This function returns 1 if the piddle has zero elements. This is useful in particular when using the indexing function which. In the case of no match to a specified criterion, the returned piddle has zero dimension.
pdl> $a=sequence(10)
pdl> $i=which($a < 1)
pdl> print "I found no matches!\n" if ($i>isempty);
I found no matches!
Note that having zero elements is rather different from the concept of being a null piddle, see the PDL::FAQ and PDL::Indexing manpages for discussions of this.
zeroes
construct a zero filled piddle from dimension list or template piddle.
Various forms of usage,
(i) by specification or (ii) by template piddle:
# usage type (i):
$a = zeroes([type], $nx, $ny, $nz,...);
$a = PDL>zeroes([type], $nx, $ny, $nz,...);
$a = $pdl>zeroes([type], $nx, $ny, $nz,...);
# usage type (ii):
$a = zeroes $b;
$a = $b>zeroes
zeroes inplace $a; # Equivalent to $a .= 0;
$a>inplace>zeroes; # ""
pdl> $z = zeroes 4,3
pdl> p $z
[
[0 0 0 0]
[0 0 0 0]
[0 0 0 0]
]
pdl> $z = zeroes ushort, 3,2 # Create ushort array
[ushort() etc. with no arg returns a PDL::Types token]
See also new_from_specification for details on using piddles in the dimensions list.
zeros
construct a zero filled piddle (see zeroes for usage)
ones
construct a one filled piddle
$a = ones([type], $nx, $ny, $nz,...);
etc. (see 'zeroes')
see zeroes() and add one
See also new_from_specification for details on using piddles in the dimensions list.
reshape
Change the shape (i.e. dimensions) of a piddle, preserving contents.
$x>reshape(NEWDIMS); reshape($x, NEWDIMS);
The data elements are preserved, obviously they will wrap differently and get truncated if the new array is shorter. If the new array is longer it will be zeropadded.
***Potential incompatibility with earlier versions of PDL**** If the list of NEWDIMS
is empty reshape
will just drop all dimensions of size 1 (preserving the number of elements):
$a = sequence(3,4,5);
$b = $a(1,3);
$b>reshape();
print $b>info;
PDL: Double D [5]
Dimensions of size 1 will also be dropped if reshape
is invoked with the argument 1:
$b = $a>reshape(1);
As opposed to reshape
without arguments, reshape(1)
preserves dataflow:
$a = ones(2,1,2);
$b = $a(0)>reshape(1);
$b++;
print $a;
[
[
[2 1]
]
[
[2 1]
]
]
Note: an explicit copy of slices is generally forced  this is the only way (for now) of stopping a crash if $x
is a slice. Important: Physical piddles are changed inplace!
pdl> $x = sequence(10)
pdl> reshape $x,3,4; p $x
[
[0 1 2]
[3 4 5]
[6 7 8]
[9 0 0]
]
pdl> reshape $x,5; p $x
[0 1 2 3 4]
squeeze
eliminate all singleton dimensions (dims of size 1)
$b = $a(0,0)>squeeze;
Alias for reshape(1)
. Removes all singleton dimensions and preserves dataflow. A more concise interface is provided by PDL::NiceSlice via modifiers:
use PDL::NiceSlice;
$b = $a(0,0;); # same as $a(0,0)>squeeze
flat
flatten a piddle (alias for $pdl>clump(1)
)
$srt = $pdl>flat>qsort;
Useful method to make a 1D piddle from an arbitrarily sized input piddle. Data flows back and forth as usual with slicing routines. Falls through if argument already <= 1D.
convert
Generic datatype conversion function
$y = convert($x, $newtypenum);
$y = convert $x, long
$y = convert $x, ushort
$newtype
is a type number, for convenience they are returned by long()
etc when called without arguments.
Datatype_conversions
byteshortushortlonglonglongfloatdouble (shorthands to convert datatypes)
$y = double $x; $y = ushort [1..10];
# all of the above listed shorthands behave similarly
When called with a piddle argument, they convert to the specific datatype.
When called with a numeric, list, listref, or string argument they construct a new piddle. This is a convenience to avoid having to be longwinded and say $x = long(pdl(42))
Thus one can say:
$a = float(1,2,3,4); # 1D
$a = float q[1 2 3; 4 5 6]; # 2D
$a = float([1,2,3],[4,5,6]); # 2D
$a = float([[1,2,3],[4,5,6]]); # 2D
Note the last three give identical results, and the last two are exactly equivalent  a list is automatically converted to a list reference for syntactic convenience. i.e. you can omit the outer []
When called with no arguments, these functions return a special type token. This allows syntactical sugar like:
$x = ones byte, 1000,1000;
This example creates a large piddle directly as byte datatype in order to save memory.
In order to control how undefs are handled in converting from perl lists to PDLs, one can set the variable $PDL::undefval
; see the function pdl() for more details.
pdl> p $x=sqrt float [1..10]
[1 1.41421 1.73205 2 2.23607 2.44949 2.64575 2.82843 3 3.16228]
pdl> p byte $x
[1 1 1 2 2 2 2 2 3 3]
byte
Convert to byte datatype
short
Convert to short datatype
ushort
Convert to ushort datatype
long
Convert to long datatype
longlong
Convert to longlong datatype
float
Convert to float datatype
double
Convert to double datatype
type
return the type of a piddle as a blessed type object
A convenience function for use with the piddle constructors, e.g.
$b = PDL>zeroes($a>type,$a>dims,3);
die "must be float" unless $a>type == float;
See also the discussion of the PDL::Type
class in PDL::Types. Note that the PDL::Type
objects have overloaded comparison and stringify operators so that you can compare and print types:
$a = $a>float if $a>type < float;
$t = $a>type; print "Type is $t\";
list
Convert piddle to perl list
@tmp = list $x;
Obviously this is grossly inefficient for the large datasets PDL is designed to handle. This was provided as a get out while PDL matured. It should now be mostly superseded by superior constructs, such as PP/threading. However it is still occasionally useful and is provied for backwards compatibility.
for (list $x) {
# Do something on each value...
}
If you compile PDL with bad value support (the default), your machine's docs will also say this:
list converts any bad values into the string 'BAD'.
unpdl
Convert piddle to nested Perl array references
$arrayref = unpdl $x;
This function returns a reference to a Perl listoflists structure equivalent to the input piddle (within the limitation that while values of elements should be preserved, the detailed datatypes will not as perl itself basically has "number" data rather than byte, short, int... E.g., sum($x  pdl( $x>unpdl ))
should equal 0.
Obviously this is grossly inefficient in memory and processing for the large datasets PDL is designed to handle. Sometimes, however, you really want to move your data back to Perl, and with proper dimensionality, unlike list
.
use JSON;
my $json = encode_json unpdl $pdl;
If you compile PDL with bad value support (the default), your machine's docs will also say this:
unpdl converts any bad values into the string 'BAD'.
listindices
Convert piddle indices to perl list
@tmp = listindices $x;
@tmp
now contains the values 0..nelem($x)
.
Obviously this is grossly inefficient for the large datasets PDL is designed to handle. This was provided as a get out while PDL matured. It should now be mostly superseded by superior constructs, such as PP/threading. However it is still occasionally useful and is provied for backwards compatibility.
for $i (listindices $x) {
# Do something on each value...
}
set
Set a single value inside a piddle
set $piddle, @position, $value
@position
is a coordinate list, of size equal to the number of dimensions in the piddle. Occasionally useful, mainly provided for backwards compatibility as superseded by use of slice and assigment operator .=
.
pdl> $x = sequence 3,4
pdl> set $x, 2,1,99
pdl> p $x
[
[ 0 1 2]
[ 3 4 99]
[ 6 7 8]
[ 9 10 11]
]
at
Returns a single value inside a piddle as perl scalar.
$z = at($piddle, @position); $z=$piddle>at(@position);
@position
is a coordinate list, of size equal to the number of dimensions in the piddle. Occasionally useful in a general context, quite useful too inside PDL internals.
pdl> $x = sequence 3,4
pdl> p $x>at(1,2)
7
If you compile PDL with bad value support (the default), your machine's docs will also say this:
at converts any bad values into the string 'BAD'.
sclr
return a single value from a piddle as a scalar
$val = $a(10)>sclr;
$val = sclr inner($a,$b);
The sclr
method is useful to turn a piddle into a normal Perl scalar. Its main advantage over using at
for this purpose is the fact that you do not need to worry if the piddle is 0D, 1D or higher dimensional. Using at
you have to supply the correct number of zeroes, e.g.
$a = sequence(10);
$b = $a>slice('4');
print $b>sclr; # no problem
print $b>at(); # error: needs at least one zero
sclr
is generally used when a Perl scalar is required instead of a oneelement piddle. If the input is a multielement piddle the first value is returned as a Perl scalar. You can optionally switch on checks to ensure that the input piddle has only one element:
PDL>sclr({Check => 'warn'}); # carp if called with multiel pdls
PDL>sclr({Check => 'barf'}); # croak if called with multiel pdls
are the commands to switch on warnings or raise an error if a multielement piddle is passed as input. Note that these options can only be set when sclr
is called as a class method (see example above). Use
PDL>sclr({Check=>0});
to switch these checks off again (default setting); When called as a class method the resulting check mode is returned (0: no checking, 1: warn, 2: barf).
cat
concatenate piddles to N+1 dimensional piddle
Takes a list of N piddles of same shape as argument, returns a single piddle of dimension N+1
pdl> $x = cat ones(3,3),zeroes(3,3),rvals(3,3); p $x
[
[
[1 1 1]
[1 1 1]
[1 1 1]
]
[
[0 0 0]
[0 0 0]
[0 0 0]
]
[
[1 1 1]
[1 0 1]
[1 1 1]
]
]
If you compile PDL with bad value support (the default), your machine's docs will also say this:
The output piddle is set bad if any input piddles have their bad flag set.
Similar functions include append and glue.
dog
Opposite of 'cat' :). Split N dim piddle to list of N1 dim piddles
Takes a single Ndimensional piddle and splits it into a list of N1 dimensional piddles. The breakup is done along the last dimension. Note the dataflown connection is still preserved by default, e.g.:
pdl> $p = ones 3,3,3
pdl> ($a,$b,$c) = dog $p
pdl> $b++; p $p
[
[
[1 1 1]
[1 1 1]
[1 1 1]
]
[
[2 2 2]
[2 2 2]
[2 2 2]
]
[
[1 1 1]
[1 1 1]
[1 1 1]
]
]
Break => 1 Break dataflow connection (new copy)
If you compile PDL with bad value support (the default), your machine's docs will also say this:
The output piddles are set bad if the original piddle has its bad flag set.
gethdr
Retrieve header information from a piddle
$pdl=rfits('file.fits');
$h=$pdl>gethdr;
print "Number of pixels in the Xdirection=$$h{NAXIS1}\n";
The gethdr
function retrieves whatever header information is contained within a piddle. The header can be set with sethdr and is always a hash reference or undef.
gethdr
returns undef if the piddle has not yet had a header defined; compare with hdr
and fhdr
, which are guaranteed to return a defined value.
Note that gethdr() works by reference: you can modify the header inplace once it has been retrieved:
$a = rfits($filename);
$ah = $a>gethdr();
$ah>{FILENAME} = $filename;
It is also important to realise that in most cases the header is not automatically copied when you copy the piddle. See hdrcpy to enable automatic header copying.
Here's another example: a wrapper around rcols that allows your piddle to remember the file it was read from and the columns could be easily written (here assuming that no regexp is needed, extensions are left as an exercise for the reader)
sub ext_rcols {
my ($file, @columns)=@_;
my $header={};
$$header{File}=$file;
$$header{Columns}=\@columns;
@piddles=rcols $file, @columns;
foreach (@piddles) { $_>sethdr($header); }
return @piddles;
}
hdr
Retrieve or set header information from a piddle
$pdl>hdr>{CDELT1} = 1;
The hdr
function allows convenient access to the header of a piddle. Unlike gethdr
it is guaranteed to return a defined value, so you can use it in a hash dereference as in the example. If the header does not yet exist, it gets autogenerated as an empty hash.
Note that this is usually  but not always  What You Want. If you want to use a tied Astro::FITS::Header hash, for example, you should either construct it yourself and use sethdr
to put it into the piddle, or use fhdr instead. (Note that you should be able to write out the FITS file successfully regardless of whether your PDL has a tied FITS header object or a vanilla hash).
fhdr
Retrieve or set FITS header information from a piddle
$pdl>fhdr>{CDELT1} = 1;
The fhdr
function allows convenient access to the header of a piddle. Unlike gethdr
it is guaranteed to return a defined value, so you can use it in a hash dereference as in the example. If the header does not yet exist, it gets autogenerated as a tied Astro::FITS::Header hash.
Astro::FITS::Header tied hashes are better at matching the behavior of FITS headers than are regular hashes. In particular, the hash keys are CAsE INsEnSItiVE, unlike normal hash keys. See Astro::FITS::Header for details.
If you do not have Astro::FITS::Header installed, you get back a normal hash instead of a tied object.
sethdr
Set header information of a piddle
$pdl = zeroes(100,100);
$h = {NAXIS=>2, NAXIS1=>100, NAXIS=>100, COMMENT=>"Sample FITSstyle header"};
# add a FILENAME field to the header
$$h{FILENAME} = 'file.fits';
$pdl>sethdr( $h );
The sethdr
function sets the header information for a piddle. You must feed in a hash ref or undef, and the header field of the PDL is set to be a new ref to the same hash (or undefined).
The hash ref requirement is a speed bump put in place since the normal use of headers is to store fits header information and the like. Of course, if you want you can hang whatever ugly old data structure you want off of the header, but that makes life more complex.
Remember that the hash is not copied  the header is made into a ref that points to the same underlying data. To get a real copy without making any assumptions about the underlying data structure, you can use one of the following:
use PDL::IO::Dumper;
$pdl>sethdr( deep_copy($h) );
(which is slow but general), or
$pdl>sethdr( PDL::_hdr_copy($h) )
(which uses the builtin sleazy deep copier), or (if you know that all the elements happen to be scalars):
{ my %a = %$h;
$pdl>sethdr(\%a);
}
which is considerably faster but just copies the top level.
The sethdr
function must be given a hash reference or undef. For further information on the header, see gethdr, hdr, fhdr and hdrcpy.
hdrcpy
switch on/off/examine automatic header copying
print "hdrs will be copied" if $a>hdrcpy;
$a>hdrcpy(1); # switch on automatic header copying
$b = $a>sumover; # and $b will inherit $a's hdr
$a>hdrcpy(0); # and now make $a noninfectious again
hdrcpy
without an argument just returns the current setting of the flag. See also "hcpy" which returns its PDL argument (and so is useful in methodcall pipelines).
Normally, the optional header of a piddle is not copied automatically in pdl operations. Switching on the hdrcpy flag using the hdrcpy
method will enable automatic hdr copying. Note that an actual deep copy gets made, which is rather processorinefficient  so avoid using header copying in tight loops!
Most PDLs have the hdrcpy
flag cleared by default; however, some routines (notably rfits) set it by default where that makes more sense.
The hdrcpy
flag is viral: if you set it for a PDL, then derived PDLs will get copies of the header and will also have their hdrcpy
flags set. For example:
$a = xvals(50,50);
$a>hdrcpy(1);
$a>hdr>{FOO} = "bar";
$b = $a++;
$c = $b++;
print $b>hdr>{FOO}, "  ", $c>hdr>{FOO}, "\n";
$b>hdr>{FOO} = "baz";
print $a>hdr>{FOO}, "  ", $b>hdr>{FOO}, "  ", $c>hdr>{FOO}, "\n";
will print:
bar  bar
bar  baz  bar
Performing an operation in which more than one PDL has its hdrcpy flag causes the resulting PDL to take the header of the first PDL:
($a,$b) = sequence(5,2)>dog;
$a>hdrcpy(1); $b>hdrcpy(1);
$a>hdr>{foo} = 'a';
$b>hdr>{foo} = 'b';
print (($a+$b)>hdr>{foo} , ($b+$a)>hdr>{foo});
will print:
a b
hcpy
Switch on/off automatic header copying, with PDL passthrough
$a = rfits('foo.fits')>hcpy(0);
$a = rfits('foo.fits')>hcpy(1);
hcpy
sets or clears the hdrcpy flag of a PDL, and returns the PDL itself. That makes it convenient for inline use in expressions.
set_autopthread_targ
Set the target number of processor threads (pthreads) for multithreaded processing.
set_autopthread_targ($num_pthreads);
$num_pthreads
is the target number of pthreads the autopthread process will try to achieve.
See PDL::ParallelCPU for an overview of the autopthread process.
# Example turning on autopthreading for a target of 2 pthreads and for functions involving
# PDLs with greater than 1M elements
set_autopthread_targ(2);
set_autopthread_size(1);
# Execute a pdl function, processing will split into two pthreads as long as
# one of the pdlthreaded dimensions is divisible by 2.
$a = minimum($b);
# Get the actual number of pthreads that were run.
$actual_pthread = get_autopthread_actual();
get_autopthread_targ
Get the current target number of processor threads (pthreads) for multithreaded processing.
$num_pthreads = get_autopthread_targ();
$num_pthreads
is the target number of pthreads the autopthread process will try to achieve.
See PDL::ParallelCPU for an overview of the autopthread process.
get_autopthread_actual
Get the actual number of pthreads executed for the last pdl processing function.
$autopthread_actual = get_autopthread_actual();
$autopthread_actual
is the actual number of pthreads executed for the last pdl processing function.
See PDL::ParallelCPU for an overview of the autopthread process.
set_autopthread_size
Set the minimum size (in Melements or 2^20 elements) of the largest PDL involved in a function where autopthreading will be performed. For small PDLs, it probably isn't worth starting multiple pthreads, so this function is used to define a minimum threshold where autopthreading won't be attempted.
set_autopthread_size($size);
$size
is the mimumum size, in Melements or 2^20 elements (approx 1e6 elements) for the largest PDL involved in a function.
See PDL::ParallelCPU for an overview of the autopthread process.
# Example turning on autopthreading for a target of 2 pthreads and for functions involving
# PDLs with greater than 1M elements
set_autopthread_targ(2);
set_autopthread_size(1);
# Execute a pdl function, processing will split into two pthreads as long as
# one of the pdlthreaded dimensions is divisible by 2.
$a = minimum($b);
# Get the actual number of pthreads that were run.
$actual_pthread = get_autopthread_actual();
get_autopthread_size
Get the current autopthread_size setting.
$autopthread_size = get_autopthread_size();
$autopthread_size
is the mimumum size limit for auto_pthreading to occur, in Melements or 2^20 elements (approx 1e6 elements) for the largest PDL involved in a function
See PDL::ParallelCPU for an overview of the autopthread process.
AUTHOR
Copyright (C) Karl Glazebrook (kgb@aaoepp.aao.gov.au), Tuomas J. Lukka, (lukka@husc.harvard.edu) and Christian Soeller (c.soeller@auckland.ac.nz) 1997. Modified, Craig DeForest (deforest@boulder.swri.edu) 2002. All rights reserved. There is no warranty. You are allowed to redistribute this software / documentation under certain conditions. For details, see the file COPYING in the PDL distribution. If this file is separated from the PDL distribution, the copyright notice should be included in the file.