C implementation of Python numpy multidimensional arrays

cspan.h

/*
 MIT License
 *
 * Copyright (c) 2025 Tyge Løvset
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in all
 * copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */
/*
 This is a single header, C implementation of python numpy multidimensional array slicing.
 It supports both row- and column-major layout, transposing of md spans, and flat iteration.
 
 cspan.h is also a part of the STC library, https://github.com/stclib/STC
 Documentation: https://github.com/stclib/STC/blob/master/docs/cspan_api.md
 Examples: https://github.com/stclib/STC/blob/master/docs/cspan_api.md
 
 The python line:
    slice = sp4[2][3:][1:5:2][:] # reduces rank from 4 to 3
 can be written in C as:
    Span3 slice = cspan_slice(&sp4, Span3, {2}, {3,c_END}, {1,5,2}, {c_ALL});
 
 Compile-time check that number of arguments matches type of sp4.
 Default runtime argument bounds check and that possible rank-reduction is type-correct.
 There are also optimized common slicing constructors (for dimensions 1-4), e.g.:
    Span sub = cspan_subspan(&sp, 5, 3);  // cspan_slice(&sp, Span, {5,8});
    Span3 ms3 = cspan_submd4(&sp3, 1);    // cspan_slice(&sp3, Span3, {1}, {c_ALL}, {c_ALL});
    Span2 ms2 = cspan_submd4(&sp3, 1, 0); // cspan_slice(&sp3, Span2, {1}, {0}, {c_ALL});
 The code generated and the implementation is fast and compact, while being
 generic and type-safe. For comparison, C++23 std::mdspan does not support
 slicing, subspan, submd, or flattened iteration.

// Example:
#include <stdio.h>
#include <stdlib.h>
#include "cspan.h"

use_cspan3(DSpan, double);

int main(void) {
    enum{nx=3, ny=4, nz=5};
    double data[nx*ny*nz];

    printf("\nMultidim span ms[3, 4, 5], fortran ordered");
    DSpan3 ms = cspan_md_layout(c_COLMAJOR, data, nx, ny, nz);

    int idx = 0;
    for (int i = 0; i < ms.shape[0]; ++i)
        for (int j = 0; j < ms.shape[1]; ++j)
            for (int k = 0; k < ms.shape[2]; ++k)
                *cspan_at(&ms, i, j, k) = ++idx;

    cspan_transpose(&ms);

    printf(", transposed:\n\n");
    cspan_print(DSpan3, "%.2f", ms);

    puts("Slicing:");
    printf("ms[0, :, :]\n");
    cspan_print(DSpan2, "%g", cspan_slice(&ms, DSpan2, {0}, {c_ALL}, {c_ALL}));

    printf("ms[:, 0, :]\n");
    cspan_print(DSpan2, "%g", cspan_slice(&ms, DSpan2, {c_ALL}, {0}, {c_ALL}));

    printf("ms[:, :, 0]\n");
    cspan_print(DSpan2, "%g", cspan_slice(&ms, DSpan2, {c_ALL}, {c_ALL}, {0}));
}
*/
#ifndef CSPAN_H_INCLUDED
#define CSPAN_H_INCLUDED

#include <assert.h>
#include <stddef.h>
#include <string.h>
#include <stdint.h>
#include <stdbool.h>

#define c_MACRO_OVERLOAD(name, ...) \
    c_JOIN(c_JOIN0(name,_), c_NUMARGS(__VA_ARGS__))(__VA_ARGS__)
#define c_TUPLE_AT_1(x,y,...) y
#define c_JOIN0(a, b) a ## b
#define c_JOIN(a, b) c_JOIN0(a, b)
#define c_EXPAND(...) __VA_ARGS__
#define c_NUMARGS(...) _c_APPLY_ARG_N((__VA_ARGS__, _c_RSEQ_N))
#define _c_APPLY_ARG_N(args) c_EXPAND(_c_ARG_N args)
#define _c_RSEQ_N 16, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0
#define _c_ARG_N(_1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, \
                 _14, _15, _16, N, ...) N

#define c_static_assert(expr)   (void)sizeof(int[(expr) ? 1 : -1])
#define c_countof(a) (intptr_t)(sizeof(a)/sizeof 0[a])
#define c_uless(a, b) ((size_t)(a) < (size_t)(b))
#define c_swap(T, xp, yp) do { T _t = *(xp); *(xp) = *(yp); *(yp) = _t; } while (0)
#define c_make_array(T, ...) ((T[])__VA_ARGS__)
#define c_make_array2d(T, N, ...) ((T[][N])__VA_ARGS__)
#define c_each(it, C, cnt) \
    C##_iter it = C##_begin(&cnt); it.ref; C##_next(&it)

#ifndef CSPAN_INDEX_TYPE
    #define CSPAN_INDEX_TYPE int32_t
#endif
typedef CSPAN_INDEX_TYPE _istride;

#define using_cspan use_cspan                   // [deprecated]
#define using_cspan2 use_cspan2                 // [deprecated]
#define using_cspan3 use_cspan3                 // [deprecated]
#define using_cspan2_with_eq use_cspan2_with_eq // [deprecated]
#define using_cspan3_with_eq use_cspan3_with_eq // [deprecated]

#define use_cspan(...) c_MACRO_OVERLOAD(use_cspan, __VA_ARGS__)
#define use_cspan_2(Self, T) \
    use_cspan_3(Self, T, 1); \
    static inline Self Self##_from_n(Self##_value* dataptr, intptr_t n) \
        { return (Self)cspan_from_n(dataptr, n); } \
    static inline const Self##_value* Self##_at(const Self* self, intptr_t idx) \
        { return cspan_at(self, idx); } \
    static inline Self##_value* Self##_at_mut(Self* self, intptr_t idx) \
        { return cspan_at(self, idx); } \
    struct stc_nostruct

#define use_cspan_with_eq(...) c_MACRO_OVERLOAD(use_cspan_with_eq, __VA_ARGS__)
#define use_cspan_with_eq_3(Self, T, i_eq) \
    use_cspan_with_eq_4(Self, T, i_eq, 1); \
    static inline Self Self##_from_n(Self##_value* dataptr, intptr_t n) \
        { return (Self)cspan_from_n(dataptr, n); } \
    struct stc_nostruct

#define use_cspan_3(Self, T, RANK) \
    typedef T Self##_value; \
    typedef T Self##_raw; \
    typedef struct { \
        Self##_value *data; \
        _istride shape[RANK]; \
        cspan_tuple##RANK stride; \
    } Self; \
    \
    typedef struct { \
        Self##_value *ref; \
        const Self *_s; \
        _istride pos[RANK]; \
    } Self##_iter; \
    \
    static inline Self Self##_slice_(Self##_value* d, const _istride shape[], const _istride stri[], \
                                  const intptr_t args[][3], const int rank) { \
        Self s; int outrank; \
        s.data = d + _cspan_slice(s.shape, s.stride.d, &outrank, shape, stri, args, rank); \
        assert(outrank == RANK); \
        return s; \
    } \
    static inline Self##_iter Self##_begin(const Self* self) { \
        return (Self##_iter){ \
            .ref=RANK==1 && self->shape[0]==0 ? NULL : self->data, ._s=self}; \
    } \
    static inline Self##_iter Self##_end(const Self* self) { \
        (void)self; \
        return (Self##_iter){0}; \
    } \
    static inline void Self##_next(Self##_iter* it) { \
        intptr_t off = it->_s->stride.d[RANK - 1]; \
        bool done = _cspan_next##RANK(it->pos, it->_s->shape, it->_s->stride.d, RANK, &off); \
        if (done) it->ref = NULL; else it->ref += off; \
    } \
    static inline intptr_t Self##_size(const Self* self) \
        { return cspan_size(self); } \
    static inline Self Self##_transposed(Self sp) \
        { _cspan_transpose(sp.shape, sp.stride.d, cspan_rank(&sp)); return sp; } \
    static inline Self Self##_swapped_axes(Self sp, int ax1, int ax2) \
        { _cspan_swap_axes(sp.shape, sp.stride.d, cspan_rank(&sp), ax1, ax2); return sp; } \
    struct stc_nostruct

#define use_cspan_with_eq_4(Self, T, i_eq, RANK) \
    use_cspan_3(Self, T, RANK); \
    static inline bool Self##_eq(const Self* x, const Self* y) { \
        if (memcmp(x->shape, y->shape, sizeof x->shape) != 0) \
            return false; \
        for (Self##_iter _i = Self##_begin(x), _j = Self##_begin(y); \
             _i.ref != NULL; Self##_next(&_i), Self##_next(&_j)) \
            { if (!(i_eq(_i.ref, _j.ref))) return false; } \
        return true; \
    } \
    static inline bool Self##_equals(Self sp1, Self sp2) \
        { return Self##_eq(&sp1, &sp2); } \
    struct stc_nostruct

#define use_cspan2(Self, T) use_cspan_2(Self, T); use_cspan_3(Self##2, T, 2)
#define use_cspan3(Self, T) use_cspan2(Self, T); use_cspan_3(Self##3, T, 3)
#define use_cspan2_with_eq(Self, T, eq) use_cspan_with_eq_3(Self, T, eq); \
                                        use_cspan_with_eq_4(Self##2, T, eq, 2)
#define use_cspan3_with_eq(Self, T, eq) use_cspan2_with_eq(Self, T, eq); \
                                        use_cspan_with_eq_4(Self##3, T, eq, 3)
#define use_cspan_tuple(N) typedef struct { _istride d[N]; } cspan_tuple##N
use_cspan_tuple(1); use_cspan_tuple(2);
use_cspan_tuple(3); use_cspan_tuple(4);
use_cspan_tuple(5); use_cspan_tuple(6);
use_cspan_tuple(7); use_cspan_tuple(8);


// Construct a cspan from a pointer+size
#define cspan_from_n(dataptr, n) \
    {.data=dataptr, \
     .shape={(_istride)(n)}, \
     .stride=(cspan_tuple1){.d={1}}}

// Create a 1d-span in the local lexical scope. N must be a compile-time constant.
#define cspan_by_copy(dataptr, N) \
    cspan_from_n(memcpy((char[(N)*sizeof *(dataptr)]){0}, dataptr, (N)*sizeof *(dataptr)), N)

// Create a zeroed out 1d-span in the local lexical scope. N must be a compile-time constant.
#define cspan_zeros(Span, N) \
    ((Span)cspan_from_n((Span##_value[N]){0}, N))

// Create a global scope 1d-span from constant initializer list, otherwise like c_make(Span, ...).
#define cspan_make(Span, ...) \
    ((Span)cspan_from_n(c_make_array(Span##_value, __VA_ARGS__), \
                        sizeof((Span##_value[])__VA_ARGS__)/sizeof(Span##_value)))

// Make 1d-span from a c-array.
#define cspan_from_array(array) \
    cspan_from_n(array, c_countof(array))

// Make 1d-span from a vec or stack container.
#define cspan_from_vec(container) \
    cspan_from_n((container)->data, (container)->size)

// Make a 1d-sub-span from a 1d-span
#define cspan_subspan(self, offset, count) \
    {.data=cspan_at(self, offset), \
     .shape={(_istride)(count)}, \
     .stride=(self)->stride}

// Accessors
//
#define cspan_size(self) _cspan_size((self)->shape, cspan_rank(self))
#define cspan_rank(self) c_countof((self)->shape) // constexpr
#define cspan_at(self, ...) ((self)->data + cspan_index(self, __VA_ARGS__))
#define cspan_front(self) ((self)->data)
#define cspan_back(self) ((self)->data + cspan_size(self) - 1)

#define cspan_index(...) cspan_index_fn(__VA_ARGS__, c_COMMA_N(cspan_index_3d), c_COMMA_N(cspan_index_2d), \
                                                     c_COMMA_N(cspan_index_1d),)(__VA_ARGS__)
#define cspan_index_fn(self, i,j,k,n, ...) c_TUPLE_AT_1(n, cspan_index_nd,)
#define cspan_index_1d(self, i)     (c_static_assert(cspan_rank(self) == 1), \
                                     assert((i) < (self)->shape[0]), \
                                     (i)*(self)->stride.d[0])
#define cspan_index_2d(self, i,j)   (c_static_assert(cspan_rank(self) == 2), \
                                     assert((i) < (self)->shape[0] && (j) < (self)->shape[1]), \
                                     (i)*(self)->stride.d[0] + (j)*(self)->stride.d[1])
#define cspan_index_3d(self, i,j,k) (c_static_assert(cspan_rank(self) == 3), \
                                     assert((i) < (self)->shape[0] && (j) < (self)->shape[1] && (k) < (self)->shape[2]), \
                                     (i)*(self)->stride.d[0] + (j)*(self)->stride.d[1] + (k)*(self)->stride.d[2])
#define cspan_index_nd(self, ...) _cspan_index((self)->shape, (self)->stride.d, c_make_array(intptr_t, {__VA_ARGS__}), \
                                               (c_static_assert(cspan_rank(self) == c_NUMARGS(__VA_ARGS__)), cspan_rank(self)))


// Multi-dimensional span constructors
//
typedef enum {c_ROWMAJOR, c_COLMAJOR, c_STRIDED} cspan_layout;

#define cspan_is_colmajor(self) \
    _cspan_is_layout(c_COLMAJOR, (self)->shape, (self)->stride.d, cspan_rank(self))
#define cspan_is_rowmajor(self) \
    _cspan_is_layout(c_ROWMAJOR, (self)->shape, (self)->stride.d, cspan_rank(self))
#define cspan_get_layout(self) \
    (cspan_is_rowmajor(self) ? c_ROWMAJOR : cspan_is_colmajor(self) ? c_COLMAJOR : c_STRIDED)

#define cspan_md(dataptr, ...) \
    cspan_md_layout(c_ROWMAJOR, dataptr, __VA_ARGS__)

// Span2 sp1 = cspan_md(data, 30, 50);
// Span2 sp2 = {data, cspan_shape(15, 25), cspan_strides(50*2, 2)}; // every second in each dim
#define cspan_shape(...) {__VA_ARGS__}
#define cspan_strides(...) {.d={__VA_ARGS__}}

#define cspan_md_layout(layout, dataptr, ...) \
    {.data=dataptr, \
     .shape={__VA_ARGS__}, \
     .stride=*(c_JOIN(cspan_tuple,c_NUMARGS(__VA_ARGS__))*) \
             _cspan_shape2stride(layout, c_make_array(_istride, {__VA_ARGS__}), c_NUMARGS(__VA_ARGS__))}

// Transpose matrix
#define cspan_transpose(self) \
    _cspan_transpose((self)->shape, (self)->stride.d, cspan_rank(self))

// Swap two matrix axes
#define cspan_swap_axes(self, ax1, ax2) \
    _cspan_swap_axes((self)->shape, (self)->stride.d, cspan_rank(self), ax1, ax2)

// Set all span elements to value.
#define cspan_set_all(Span, self, value) do { \
    Span##_value _v = value; \
    for (c_each(_it, Span, *(self))) *_it.ref = _v; \
} while (0)

// General slicing function.
//
#define c_END (_istride)(((size_t)1 << (sizeof(_istride)*8 - 1)) - 1)
#define c_ALL 0,c_END

#define cspan_slice(self, Outspan, ...) \
    Outspan##_slice_((self)->data, (self)->shape, (self)->stride.d, \
                     c_make_array2d(const intptr_t, 3, {__VA_ARGS__}), \
                     (c_static_assert(cspan_rank(self) == sizeof((intptr_t[][3]){__VA_ARGS__})/sizeof(intptr_t[3])), cspan_rank(self)))

// submd#(): Reduces rank, fully typesafe + range checked by default
//           int ms3[N1][N2][N3];
//           int (*ms2)[N3] = ms3[1]; // traditional, lose range test/info. VLA.
//           Span3 ms3 = cspan_md(data, N1,N2,N3); // Uses cspan_md instead.
//           *cspan_at(&ms3, 1,1,1) = 42;
//           Span2 ms2 = cspan_slice(&ms3, Span2, {1}, {c_ALL}, {c_ALL});
//           Span2 ms2 = cspan_submd3(&ms3, 1); // Same as line above, optimized.
#define cspan_submd2(self, x) \
    {.data=cspan_at(self, x, 0), \
     .shape={(self)->shape[1]}, \
     .stride=(cspan_tuple1){.d={(self)->stride.d[1]}}}

#define cspan_submd3(...) c_MACRO_OVERLOAD(cspan_submd3, __VA_ARGS__)
#define cspan_submd3_2(self, x) \
    {.data=cspan_at(self, x, 0, 0), \
     .shape={(self)->shape[1], (self)->shape[2]}, \
     .stride=(cspan_tuple2){.d={(self)->stride.d[1], (self)->stride.d[2]}}}
#define cspan_submd3_3(self, x, y) \
    {.data=cspan_at(self, x, y, 0), \
     .shape={(self)->shape[2]}, \
     .stride=(cspan_tuple1){.d={(self)->stride.d[2]}}}

#define cspan_submd4(...) c_MACRO_OVERLOAD(cspan_submd4, __VA_ARGS__)
#define cspan_submd4_2(self, x) \
    {.data=cspan_at(self, x, 0, 0, 0), \
     .shape={(self)->shape[1], (self)->shape[2], (self)->shape[3]}, \
     .stride=(cspan_tuple3){.d={(self)->stride.d[1], (self)->stride.d[2], (self)->stride.d[3]}}}
#define cspan_submd4_3(self, x, y) \
    {.data=cspan_at(self, x, y, 0, 0), \
     .shape={(self)->shape[2], (self)->shape[3]}, \
     .stride=(cspan_tuple2){.d={(self)->stride.d[2], (self)->stride.d[3]}}}
#define cspan_submd4_4(self, x, y, z) \
    {.data=cspan_at(self, x, y, z, 0), \
     .shape={(self)->shape[3]}, \
     .stride=(cspan_tuple1){.d={(self)->stride.d[3]}}}

#define cspan_print(...) c_MACRO_OVERLOAD(cspan_print, __VA_ARGS__)
#define cspan_print_3(Span, fmt, span) \
    cspan_print_4(Span, fmt, span, stdout)
#define cspan_print_4(Span, fmt, span, fp) \
    cspan_print_5(Span, fmt, span, fp, "[]")
#define cspan_print_5(Span, fmt, span, fp, brackets) \
    cspan_print_6(Span, fmt, span, fp, brackets, c_EXPAND)
#define cspan_print_complex(Span, prec, span, fp) \
    cspan_print_6(Span, "%." #prec "f%+." #prec "fi", span, fp, "[]", cspan_CMPLX_FLD)
#define cspan_CMPLX_FLD(x) creal(x), cimag(x)

#define cspan_print_6(Span, fmt, span, fp, brackets, field) do { \
    const Span _s = span; \
    const char *_f = fmt, *_b = brackets; \
    FILE* _fp = fp; \
    int _w, _max = 0; \
    char _res[2][20], _fld[64]; \
    for (c_each(_it, Span, _s)) { \
        _w = snprintf(NULL, 0ULL, _f, field(_it.ref[0])); \
        if (_w > _max) _max = _w; \
    } \
    for (c_each(_it, Span, _s)) { \
        _cspan_print_assist(_it.pos, _s.shape, cspan_rank(&_s), _b, _res); \
        _w = _max + (_it.pos[cspan_rank(&_s) - 1] > 0); \
        snprintf(_fld, sizeof _fld, _f, field(_it.ref[0])); \
        fprintf(_fp, "%s%*s%s", _res[0], _w, _fld, _res[1]); \
    } \
} while (0)

/* ----- PRIVATE ----- */

static inline intptr_t _cspan_size(const _istride shape[], int rank) {
    intptr_t size = shape[0];
    while (--rank) size *= shape[rank];
    return size;
}

static inline void _cspan_swap_axes(_istride shape[], _istride stride[],
                                    int rank, int ax1, int ax2) {
    (void)rank;
    assert(c_uless(ax1, rank) & c_uless(ax2, rank));
    c_swap(_istride, shape + ax1, shape + ax2);
    c_swap(_istride, stride + ax1, stride + ax2);
}

static inline void _cspan_transpose(_istride shape[], _istride stride[], int rank) {
    for (int i = 0; i < --rank; ++i) {
        c_swap(_istride, shape + i, shape + rank);
        c_swap(_istride, stride + i, stride + rank);
    }
}

static inline intptr_t _cspan_index(const _istride shape[], const _istride stride[],
                                    const intptr_t args[], int rank) {
    intptr_t off = 0;
    (void)shape;
    while (rank-- != 0) {
        assert(args[rank] < shape[rank]);
        off += args[rank]*stride[rank];
    }
    return off;
}

void _cspan_print_assist(_istride pos[], const _istride shape[], const int rank,
                         const char* brackets, char result[2][20]);

bool _cspan_nextN(_istride pos[], const _istride shape[], const _istride stride[], int rank, intptr_t* off);
#define _cspan_next1(pos, shape, stride, rank, off)            (++pos[0] == shape[0])
#define _cspan_next2(pos, shape, stride, rank, off)            (++pos[1] == shape[1] && \
    (pos[1] = 0, *off += stride[0] - (intptr_t)shape[1]*stride[1], ++pos[0] == shape[0]))
#define _cspan_next3(pos, shape, stride, rank, off)            (++pos[2] == shape[2] && \
    (pos[2] = 0, *off += stride[1] - (intptr_t)shape[2]*stride[2], ++pos[1] == shape[1]) && \
    (pos[1] = 0, *off += stride[0] - (intptr_t)shape[1]*stride[1], ++pos[0] == shape[0]))
#define _cspan_next4 _cspan_nextN
#define _cspan_next5 _cspan_nextN
#define _cspan_next6 _cspan_nextN
#define _cspan_next7 _cspan_nextN
#define _cspan_next8 _cspan_nextN

intptr_t _cspan_slice(_istride oshape[], _istride ostride[], int* orank,
                      const _istride shape[], const _istride stride[],
                      const intptr_t args[][3], int rank);
_istride* _cspan_shape2stride(cspan_layout layout, _istride shape[], int rank);
bool _cspan_is_layout(cspan_layout layout, const _istride shape[], const _istride strides[], int rank);

#endif // CSPAN_H_INCLUDED

/* --------------------- IMPLEMENTATION --------------------- */
#ifdef CSPAN_IMPLEMENT

bool _cspan_is_layout(cspan_layout layout, const _istride shape[], const _istride strides[], int rank) {
    _istride tmpshape[16]; // 16 = "max" rank
    size_t sz = (size_t)rank*sizeof(_istride);
    memcpy(tmpshape, shape, sz);
    return memcmp(strides, _cspan_shape2stride(layout, tmpshape, rank), sz) == 0;
}

void _cspan_print_assist(_istride pos[], const _istride shape[], const int rank,
                         const char* brackets, char result[2][20]) {
    int n = 0, j = 0, r = rank - 1;
    memset(result, 0, 32);

    // left braces:
    while (n <= r && pos[r - n] == 0)
        ++n;
    if (n) for (; j < rank; ++j)
        result[0][j] = j < rank - n ? ' ' : brackets[0];

    // right braces:
    for (j = 0; r >= 0 && pos[r] + 1 == shape[r]; --r, ++j)
        result[1][j] = brackets[1];

    // comma and newlines:
    n = (j > 0) + ((j > 1) & (j < rank));
    if (brackets[2] && j < rank)
        result[1][j++] = brackets[2]; // comma
    while (n--)
        result[1][j++] = '\n';
}

bool _cspan_nextN(_istride pos[], const _istride shape[], const _istride stride[],
                  int rank, intptr_t* off) {
    ++pos[--rank];
    for (; rank && pos[rank] == shape[rank]; --rank) {
        pos[rank] = 0; ++pos[rank - 1];
        *off += stride[rank - 1] - (intptr_t)shape[rank]*stride[rank];
    }
    return pos[rank] == shape[rank];
}

_istride* _cspan_shape2stride(cspan_layout layout, _istride shpstri[], int rank) {
    int i, inc;
    if (layout == c_COLMAJOR) i = 0, inc = 1;
    else i = rank - 1, inc = -1;
    _istride k = 1, s1 = shpstri[i], s2;

    shpstri[i] = 1;
    while (--rank) {
        i += inc;
        s2 = shpstri[i];
        shpstri[i] = (k *= s1);
        s1 = s2;
    }
    return shpstri;
}

intptr_t _cspan_slice(_istride oshape[], _istride ostride[], int* orank,
                      const _istride shape[], const _istride stride[],
                      const intptr_t args[][3], int rank) {
    intptr_t end, off = 0;
    int i = 0, oi = 0;

    for (; i < rank; ++i) {
        off += args[i][0]*stride[i];
        switch (args[i][1]) {
            case 0: assert(c_uless(args[i][0], shape[i])); continue;
            case c_END: end = shape[i]; break;
            default: end = args[i][1];
        }
        oshape[oi] = (_istride)(end - args[i][0]);
        ostride[oi] = stride[i];
        assert((oshape[oi] > 0) & !c_uless(shape[i], end));
        if (args[i][2] > 0) {
            ostride[oi] *= (_istride)args[i][2];
            oshape[oi] = (oshape[oi] - 1)/(_istride)args[i][2] + 1;
        }
        ++oi;
    }
    *orank = oi;
    return off;
}
#endif // CSPAN_IMPLEMENT

Good question. I need It only to write the cspan_md() / cspan_md_layout() constructor macros in this generic way. It returns a cspan_tupleN type, where N depends on number of args passed to it. I could have avoided wrapping the stride by making one cspan_mdN() macro per dimension, which would have made access to the stride a bit more natural syntactically.

I am new to this C macro trickery. What is the limitation that prevents you from doing this for a general size?

Could you use another macro to dispatch to cspan_md1, cspan_md2, etc. based on the number of arguments?

BTW, this is a very cool thing that you have accomplished here.

As a reminder for myself, the cspan_submdN() functions does not work for the "general size" as they are only defined for dims 2, 3, 4. However, these are just convenience/optimized versions of the general cspan_slice() function.

Beyond 14 dimensions, you would need to extend the _c_ARG_N and _c_RSEQ_N.
Beyond 8 dimensions you would need to (e.g. for dim 9):

#define  _cspan_next9 _cspan_nextN
use_cspan_tuple(9);