Changes between Version 25 and Version 26 of SwiggingHlsvd

Timestamp:

Jul 14, 2009, 5:03:48 PM (10 years ago)

Author:

flip

Comment:

--

SwiggingHlsvd

@@ -184 +184 @@
 {{{
 #!python
-    yr = hlsvd_wrapper.new_float_array(2048)
-    yi = hlsvd_wrapper.new_float_array(2048)
-    ndata = hlsvd_wrapper.new_ndata()
-    ndp = hlsvd_wrapper.new_ndp()
-    tstep = hlsvd_wrapper.new_tstep()
-    m = hlsvd_wrapper.new_m()
-    nssv = hlsvd_wrapper.new_nssv()
-    nit = hlsvd_wrapper.new_nit()
-    rootr = hlsvd_wrapper.new_double_array(50)
-    rooti = hlsvd_wrapper.new_double_array(50)
-    fre = hlsvd_wrapper.new_float_array(50)
-    dam = hlsvd_wrapper.new_float_array(50)
-    amp = hlsvd_wrapper.new_float_array(50)
-    phase = hlsvd_wrapper.new_float_array(50)
-    sv = hlsvd_wrapper.new_double_array(1024)
-    snoise = hlsvd_wrapper.new_snoise()
-    rms = hlsvd_wrapper.new_rms()
-    ndiv = hlsvd_wrapper.new_ndiv()
-    nitf = hlsvd_wrapper.new_nitf()
+yr = hlsvd_wrapper.new_float_array(2048)
+yi = hlsvd_wrapper.new_float_array(2048)
+ndata = hlsvd_wrapper.new_ndata()
+ndp = hlsvd_wrapper.new_ndp()
+tstep = hlsvd_wrapper.new_tstep()
+m = hlsvd_wrapper.new_m()
+nssv = hlsvd_wrapper.new_nssv()
+nit = hlsvd_wrapper.new_nit()
+rootr = hlsvd_wrapper.new_double_array(50)
+rooti = hlsvd_wrapper.new_double_array(50)
+fre = hlsvd_wrapper.new_float_array(50)
+dam = hlsvd_wrapper.new_float_array(50)
+amp = hlsvd_wrapper.new_float_array(50)
+phase = hlsvd_wrapper.new_float_array(50)
+sv = hlsvd_wrapper.new_double_array(1024)
+snoise = hlsvd_wrapper.new_snoise()
+rms = hlsvd_wrapper.new_rms()
+ndiv = hlsvd_wrapper.new_ndiv()
+nitf = hlsvd_wrapper.new_nitf()
 }}}
 
@@ -209 +209 @@
 {{{
 #!python
-    hlsvd_wrapper.hlsvd(yr, yi, ndata, ndp, tstep, m, nssv, 
-                         rootr, rooti, fre, dam, amp, phase, sv, snoise, 
-                         rms, ndiv, nit, nitf)
-}}}
-
+hlsvd_wrapper.hlsvd(yr, yi, ndata, ndp, tstep, m, nssv, 
+                    rootr, rooti, fre, dam, amp, phase, sv, snoise, 
+                    rms, ndiv, nit, nitf)
+}}}
+
+{{{
+#!rst
 Note that every disadvantage of the C interface remains. There's no distinction between input and output parameters. The complex numbers are still represented as correlated lists of floats. The list/array sizes must be specified explicitly, and an opportunity for memory corruption or a crash exists if one gets it wrong. One must keep track of which variables are floats and which are doubles. And it's many more lines of code than the Python version. (Some of this is exacerbated by ``hlsvd()``'s pass-by-reference interface.)
 
@@ -233 +235 @@
 
 One starts with the very unPythonic explicit load of the C library to be called:
-::
-
-    lib = ctypes.CDLL("hlsvd.dylib")
-
+}}}
+
+{{{
+#!python
+lib = ctypes.CDLL("hlsvd.dylib")
+}}}
 
 This is one place where SWIG is more Pythonic, since I access SWIG's library with a normal Python import statement. 
@@ -242 +246 @@
 
 Ctypes already knows about, well, C types, so it's easy to create something that can be passed to a function expecting a C float, for instance:
-::
-
-   tstep = ctypes.c_float(.256)
-
+
+{{{
+#!python
+tstep = ctypes.c_float(.256)
+}}}
 
 Arrays are pretty simple too:
-::
-
-   # Create an array type; this is like a class
-   InputFloatArrayType = ctypes.c_float * 2048
-   # Create an instance of that "class"
-   yr = InputFloatArrayType()
-
+
+{{{
+#!python
+# Create an array type; this is like a class
+InputFloatArrayType = ctypes.c_float * 2048
+# Create an instance of that "class"
+yr = InputFloatArrayType()
+}}}
 
 The variable setup for the call to ``hlsvd()`` looks like the code below:
-::
-
-    InputFloatArrayType = ctypes.c_float * 2048
-    OutputDoubleArrayType = ctypes.c_double * 50
-    OutputFloatArrayType = ctypes.c_float * 50
-
-    yr = InputFloatArrayType()
-    yi = InputFloatArrayType()
-
-    ndpmax = ctypes.c_long(2048)
-    ndp = ctypes.c_long(1024)
-    tstep = ctypes.c_float(.256)
-    m = ctypes.c_long(380)
-    nssv = ctypes.c_long(25)
-    nit = ctypes.c_long(500)
-
-    rootr = OutputDoubleArrayType()
-    rooti = OutputDoubleArrayType()
-    fre = OutputFloatArrayType()
-    dam = OutputFloatArrayType()
-    amp = OutputFloatArrayType()
-    phase = OutputFloatArrayType()
-    sv = OutputDoubleArrayType()
-    snoise = ctypes.c_double()
-    rms = ctypes.c_double()
-    ndiv = ctypes.c_long()
-    nitf = ctypes.c_long()
-
-
-    lib.hlsvd(yr, yi, ctypes.byref(ndpmax), ctypes.byref(ndp),
-               ctypes.byref(tstep), ctypes.byref(m), ctypes.byref(nssv),
-               ctypes.pointer(rootr), ctypes.pointer(rooti), 
-               ctypes.pointer(fre), ctypes.pointer(dam), 
-               ctypes.pointer(amp), ctypes.pointer(phase), 
-               ctypes.pointer(sv), ctypes.pointer(snoise), 
-               ctypes.pointer(rms), ctypes.pointer(ndiv),
-               ctypes.byref(nit), ctypes.pointer(nitf)
-              )
-
+
+{{{
+#!python
+InputFloatArrayType = ctypes.c_float * 2048
+OutputDoubleArrayType = ctypes.c_double * 50
+OutputFloatArrayType = ctypes.c_float * 50
+
+yr = InputFloatArrayType()
+yi = InputFloatArrayType()
+
+ndpmax = ctypes.c_long(2048)
+ndp = ctypes.c_long(1024)
+tstep = ctypes.c_float(.256)
+m = ctypes.c_long(380)
+nssv = ctypes.c_long(25)
+nit = ctypes.c_long(500)
+
+rootr = OutputDoubleArrayType()
+rooti = OutputDoubleArrayType()
+fre = OutputFloatArrayType()
+dam = OutputFloatArrayType()
+amp = OutputFloatArrayType()
+phase = OutputFloatArrayType()
+sv = OutputDoubleArrayType()
+snoise = ctypes.c_double()
+rms = ctypes.c_double()
+ndiv = ctypes.c_long()
+nitf = ctypes.c_long()
+
+
+lib.hlsvd(yr, yi, ctypes.byref(ndpmax), ctypes.byref(ndp),
+           ctypes.byref(tstep), ctypes.byref(m), ctypes.byref(nssv),
+           ctypes.pointer(rootr), ctypes.pointer(rooti), 
+           ctypes.pointer(fre), ctypes.pointer(dam), 
+           ctypes.pointer(amp), ctypes.pointer(phase), 
+           ctypes.pointer(sv), ctypes.pointer(snoise), 
+           ctypes.pointer(rms), ctypes.pointer(ndiv),
+           ctypes.byref(nit), ctypes.pointer(nitf)
+          )
+}}}
 
 At first glance this might look no better than the SWIG code, but it's actually a bit nicer. For one thing, the array sizes are defined in one place (the first three lines) and not repeated over and over. In addition, it's pretty clear what's going on in a line like this:
-::
-
-    ndiv = ctypes.c_long()
+
+{{{
+#!python
+ndiv = ctypes.c_long()
+}}}
 
 whereas SWIG's version is opaque:
-::
-
-    ndiv = hlsvd_wrapper.new_ndiv()
-
+
+{{{
+#!python
+ndiv = hlsvd_wrapper.new_ndiv()
+}}}
 
 The ``ctypes.byref()`` call is informative. It says that the C interface demands that I pass by reference, but I don't need access to the pointer here in Python. That tells anyone reading this code that I probably don't care what the called function does with the value, hinting that it is input-only. Again, I wouldn't need this at all if the f2c-generated code was tweaked to eliminate the unnecessary pass-by-reference for the input parameters.