Mercurial > octave
view liboctave/array/dim-vector.cc @ 21573:f3f8e1d3e399 stable
avoid mulitple definitions of static function-scope vars (bug #47372)
Some systems (Windows, others?) may generate multiple definitions of
function-scope static variables if those functions appear in header
files. This can cause trouble in various ways. Attempt to avoid the
problem by moving functions that contain file-scope varaibles from
header files to source files, or by using class-scope static variables
instead. Also attempt to avoid similar problems caused by implicit
instantiations of templates that have functions with static variables.
* Cell.h, Cell.cc (Cell::resize_fill_value):
Move definition from header to source file.
* oct-map.h, oct-map.cc (octave_fields::nil_rep): Likewise.
* ov-base-scalar.h, ov-base-scalar.cc (octave_base_scalar<ST>::dims):
Likewise.
* ov-fcn-handle.h, ov-fcn-handle.cc (octave_fcn_handle::dims): Likewise.
* ov.h, ov.cc (octave_value::nil_rep): Likewise.
* Array.h, Array.cc (Array<T>::nil_rep): Likewise.
* DiagArray2.h, DiagArray2.cc (DiagArray2<T>::elem,
DiagArray2<T>::checkelem): Likewise.
* Sparse.h, Sparse.cc (Sparse<T>::nil_rep): Likewise.
* dim-vector.h, dim-vector.cc (dim_vector::nil_rep): Likewise.
* idx-vector.h, idx-vector.cc (idx_vector::nil_rep,
idx_vector::err_rep): Likewise.
* oct-inttypes.h, oct-inttypes.cc (octave_int_base<T>::convert_real):
Likewise.
* symtab.h, symtab.cc (octave_value symbol_table::dummy_octave_value,
symbol_table::symbol_record symbol_table::dummy_symbol_record): New
class-scope static variables. Use them to eliminate function-scope
static variables.
* ov-int8.cc, ov-int16.cc, ov-int32.cc, ov-int64.cc, ov-uint8.cc,
ov-uint16.cc, ov-uint32.cc, ov-uint64.cc, ov-scalar.cc,
ov-flt-complex.cc, ov-complex.cc, ov-classdef.cc, ov-bool.cc,
Array-jit.cc, Array-os.cc, Array-tc.cc, Array-C.cc, Array-b.cc,
Array-ch.cc, Array-d.cc, Array-f.cc, Array-fC.cc, Array-i.cc,
Array-idx-vec.cc, Array-s.cc, Array-str.cc, Array-voidp.cc:
Prevent impliticit template instantiation.
* Array-tc.cc: Also instantiate Array<cdef_object>.
author | John W. Eaton <jwe@octave.org> |
---|---|
date | Fri, 01 Apr 2016 12:04:04 -0400 |
parents | 4197fc428c7d |
children | ae4d7dfea337 |
line wrap: on
line source
/* Copyright (C) 2003-2015 John W. Eaton Copyirght (C) 2009, 2010 VZLU Prague This file is part of Octave. Octave is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version. Octave is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with Octave; see the file COPYING. If not, see <http://www.gnu.org/licenses/>. */ #ifdef HAVE_CONFIG_H #include <config.h> #endif #include <iostream> #include "dim-vector.h" octave_idx_type * dim_vector::nil_rep (void) { static dim_vector zv (0, 0); return zv.rep; } // The maximum allowed value for a dimension extent. This will normally be a // tiny bit off the maximum value of octave_idx_type. // Currently 1 is subtracted to allow safe conversion of any 2D Array into // Sparse, but this offset may change in the future. octave_idx_type dim_vector::dim_max (void) { return std::numeric_limits<octave_idx_type>::max () - 1; } void dim_vector::chop_all_singletons (void) { make_unique (); int j = 0; int l = ndims (); for (int i = 0; i < l; i++) { if (rep[i] != 1) rep[j++] = rep[i]; } if (j == 1) rep[1] = 1; ndims () = j > 2 ? j : 2; } std::string dim_vector::str (char sep) const { std::ostringstream buf; for (int i = 0; i < length (); i++) { buf << elem (i); if (i < length () - 1) buf << sep; } std::string retval = buf.str (); return retval; } int dim_vector::num_ones (void) const { int retval = 0; for (int i = 0; i < length (); i++) if (elem (i) == 1) retval++; return retval; } octave_idx_type dim_vector::safe_numel (void) const { octave_idx_type idx_max = dim_max (); octave_idx_type n = 1; int n_dims = length (); for (int i = 0; i < n_dims; i++) { n *= rep[i]; if (rep[i] != 0) idx_max /= rep[i]; if (idx_max <= 0) throw std::bad_alloc (); } return n; } dim_vector dim_vector::squeeze (void) const { dim_vector new_dims = *this; bool dims_changed = 1; int k = 0; for (int i = 0; i < length (); i++) { if (elem (i) == 1) dims_changed = true; else new_dims(k++) = elem (i); } if (dims_changed) { if (k == 0) new_dims = dim_vector (1, 1); else if (k == 1) { // There is one non-singleton dimension, so we need // to decide the correct orientation. if (elem (0) == 1) { // The original dimension vector had a leading // singleton dimension. octave_idx_type tmp = new_dims(0); new_dims.resize (2); new_dims(0) = 1; new_dims(1) = tmp; } else { // The first element of the original dimension vector // was not a singleton dimension. new_dims.resize (2); new_dims(1) = 1; } } else new_dims.resize (k); } return new_dims; } // This is the rule for cat(). cat (dim, A, B) works if one // of the following holds, in this order: // // 1. size (A, k) == size (B, k) for all k != dim. // In this case, size (C, dim) = size (A, dim) + size (B, dim) and // other sizes remain intact. // // 2. A is 0x0, in which case B is the result // 3. B is 0x0, in which case A is the result bool dim_vector::concat (const dim_vector& dvb, int dim) { int orig_nd = ndims (); int ndb = dvb.ndims (); int new_nd = dim < ndb ? ndb : dim + 1; if (new_nd > orig_nd) resize (new_nd, 1); else new_nd = orig_nd; make_unique (); bool match = true; for (int i = 0; i < ndb; i++) { if (i != dim && rep[i] != dvb(i)) { match = false; break; } } for (int i = ndb; i < new_nd; i++) { if (i != dim && rep[i] != 1) { match = false; break; } } if (match) rep[dim] += (dim < ndb ? dvb(dim) : 1); else { // Dimensions don't match. The only allowed fix is // to omit 0x0. if (ndb == 2 && dvb(0) == 0 && dvb(1) == 0) match = true; else if (orig_nd == 2 && rep[0] == 0 && rep[1] == 0) { *this = dvb; match = true; } } chop_trailing_singletons (); return match; } // Rules for horzcat/vertcat are yet looser. // two arrays A, B can be concatenated // horizontally (dim = 2) or vertically (dim = 1) if one of the // following holds, in this order: // // 1. cat (dim, A, B) works // // 2. A, B are 2D and one of them is an empty vector, in which // case the result is the other one except if both of them // are empty vectors, in which case the result is 0x0. bool dim_vector::hvcat (const dim_vector& dvb, int dim) { if (concat (dvb, dim)) return true; else if (length () == 2 && dvb.length () == 2) { bool e2dv = rep[0] + rep[1] == 1; bool e2dvb = dvb(0) + dvb(1) == 1; if (e2dvb) { if (e2dv) *this = dim_vector (); return true; } else if (e2dv) { *this = dvb; return true; } } return false; } dim_vector dim_vector::redim (int n) const { int n_dims = length (); if (n_dims == n) return *this; else if (n_dims < n) { dim_vector retval = alloc (n); for (int i = 0; i < n_dims; i++) retval.rep[i] = rep[i]; for (int i = n_dims; i < n; i++) retval.rep[i] = 1; return retval; } else { if (n < 1) n = 1; dim_vector retval = alloc (n); retval.rep[1] = 1; for (int i = 0; i < n-1; i++) retval.rep[i] = rep[i]; int k = rep[n-1]; for (int i = n; i < n_dims; i++) k *= rep[i]; retval.rep[n-1] = k; return retval; } }