Merge pull request #59 from pfalcon/slice

Implement basic slice object and string slicing

py/mpconfig.h

@@ -25,3 +25,9 @@
 #ifndef MICROPY_MEM_STATS
 #define MICROPY_MEM_STATS (1)
 #endif
+
+// Whether to support slice object and correspondingly
+// slice subscript operators
+#ifndef MICROPY_ENABLE_SLICE
+#define MICROPY_ENABLE_SLICE (1)
+#endif

py/obj.h

@@ -144,6 +144,7 @@ mp_obj_t mp_obj_new_list(uint n, mp_obj_t *items);
 mp_obj_t mp_obj_new_list_reverse(uint n, mp_obj_t *items);
 mp_obj_t mp_obj_new_dict(int n_args);
 mp_obj_t mp_obj_new_set(int n_args, mp_obj_t *items);
+mp_obj_t mp_obj_new_slice(mp_obj_t start, mp_obj_t stop, mp_obj_t step);
 mp_obj_t mp_obj_new_bound_meth(mp_obj_t self, mp_obj_t meth);
 mp_obj_t mp_obj_new_class(struct _mp_map_t *class_locals);
 mp_obj_t mp_obj_new_instance(mp_obj_t clas);
@@ -214,6 +215,10 @@ mp_obj_t mp_obj_dict_store(mp_obj_t self_in, mp_obj_t key, mp_obj_t value);
 // set
 void mp_obj_set_store(mp_obj_t self_in, mp_obj_t item);
 
+// slice
+extern const mp_obj_type_t slice_type;
+void mp_obj_slice_get(mp_obj_t self_in, machine_int_t *start, machine_int_t *stop, machine_int_t *step);
+
 // functions
 typedef struct _mp_obj_fun_native_t { // need this so we can define const objects (to go in ROM)
     mp_obj_base_t base;

py/objslice.c

@@ -0,0 +1,74 @@
+#include <stdlib.h>
+#include <stdint.h>
+#include <string.h>
+#include <assert.h>
+
+#include "nlr.h"
+#include "misc.h"
+#include "mpconfig.h"
+#include "obj.h"
+#include "runtime0.h"
+
+#if MICROPY_ENABLE_SLICE
+
+// TODO: This implements only variant of slice with 2 integer args only.
+// CPython supports 3rd arg (step), plus args can be arbitrary Python objects.
+typedef struct _mp_obj_slice_t {
+    mp_obj_base_t base;
+    machine_int_t start;
+    machine_int_t stop;
+} mp_obj_slice_t;
+
+void slice_print(void (*print)(void *env, const char *fmt, ...), void *env, mp_obj_t o_in) {
+    mp_obj_slice_t *o = o_in;
+    print(env, "slice(" INT_FMT ", " INT_FMT ")", o->start, o->stop);
+}
+
+const mp_obj_type_t slice_type = {
+    { &mp_const_type },
+    "slice",
+    slice_print,
+    NULL, // call_n
+    NULL, // unary_op
+    NULL, // binary_op
+    NULL, // getiter
+    NULL, // iternext
+    { { NULL, NULL }, }, // method list
+};
+
+// TODO: Make sure to handle "empty" values, which are signified by None in CPython
+mp_obj_t mp_obj_new_slice(mp_obj_t ostart, mp_obj_t ostop, mp_obj_t ostep) {
+    assert(ostep == NULL);
+    machine_int_t start = 0, stop = 0;
+    if (ostart != mp_const_none) {
+        start = mp_obj_get_int(ostart);
+    }
+    if (ostop != mp_const_none) {
+        stop = mp_obj_get_int(ostop);
+        if (stop == 0) {
+            // [x:0] is a special case - in our slice object, stop = 0 means
+            // "end of sequence". Fortunately, [x:0] is an empty seqence for
+            // any x (including negative). [x:x] is also always empty sequence.
+            // but x also can be 0. But note that b""[x:x] is b"" for any x (i.e.
+            // no IndexError, at least in Python 3.3.3). So, we just use -1's to
+            // signify that. -1 is catchy "special" number in case someone will
+            // try to print [x:0] slice ever.
+            start = stop = -1;
+        }
+    }
+    mp_obj_slice_t *o = m_new(mp_obj_slice_t, 1);
+    o->base.type = &slice_type;
+    o->start = start;
+    o->stop = stop;
+    return (mp_obj_t)o;
+}
+
+void mp_obj_slice_get(mp_obj_t self_in, machine_int_t *start, machine_int_t *stop, machine_int_t *step) {
+    assert(MP_OBJ_IS_TYPE(self_in, &slice_type));
+    mp_obj_slice_t *self = self_in;
+    *start = self->start;
+    *stop = self->stop;
+    *step = 1;
+}
+
+#endif

py/objstr.c

@@ -27,9 +27,31 @@ mp_obj_t str_binary_op(int op, mp_obj_t lhs_in, mp_obj_t rhs_in) {
     const char *lhs_str = qstr_str(lhs->qstr);
     switch (op) {
         case RT_BINARY_OP_SUBSCR:
-            // string access
-            // XXX a massive hack!
-            return mp_obj_new_int(lhs_str[mp_obj_get_int(rhs_in)]);
+            // TODO: need predicate to check for int-like type (bools are such for example)
+            // ["no", "yes"][1 == 2] is common idiom
+            if (MP_OBJ_IS_SMALL_INT(rhs_in)) {
+                // TODO: This implements byte string access for single index so far
+                // TODO: Handle negative indexes.
+                return mp_obj_new_int(lhs_str[mp_obj_get_int(rhs_in)]);
+#if MICROPY_ENABLE_SLICE
+            } else if (MP_OBJ_IS_TYPE(rhs_in, &slice_type)) {
+                int start, stop, step;
+                mp_obj_slice_get(rhs_in, &start, &stop, &step);
+                assert(step == 1);
+                int len = strlen(lhs_str);
+                if (start < 0) {
+                    start = len + start;
+                }
+                if (stop <= 0) {
+                    stop = len + stop;
+                }
+                return mp_obj_new_str(qstr_from_strn_copy(lhs_str + start, stop - start));
+#endif
+            } else {
+                // Message doesn't match CPython, but we don't have so much bytes as they
+                // to spend them on verbose wording
+                nlr_jump(mp_obj_new_exception_msg(rt_q_TypeError, "index must be int"));
+            }
 
         case RT_BINARY_OP_ADD:
         case RT_BINARY_OP_INPLACE_ADD:

py/vm.c

@@ -410,6 +410,20 @@ bool mp_execute_byte_code_2(const byte **ip_in_out, mp_obj_t *fastn, mp_obj_t **
                         sp++;
                         break;
 
+#if MICROPY_ENABLE_SLICE
+                    case MP_BC_BUILD_SLICE:
+                        DECODE_UINT;
+                        if (unum == 2) {
+                            obj2 = POP();
+                            obj1 = TOP();
+                            SET_TOP(mp_obj_new_slice(obj1, obj2, NULL));
+                        } else {
+                            printf("3-argument slice is not supported\n");
+                            assert(0);
+                        }
+                        break;
+#endif
+
                     case MP_BC_UNPACK_SEQUENCE:
                         DECODE_UINT;
                         rt_unpack_sequence(sp[0], unum, sp - unum + 1);

stm/Makefile

@@ -78,6 +78,7 @@ PY_O = \
 	objnone.o \
 	objrange.o \
 	objset.o \
+	objslice.o \
 	objstr.o \
 	objtuple.o \
 	objtype.o \

tests/basics/tests/slice-bstr1.py

@@ -0,0 +1,27 @@
+b"123"[0:1]
+
+b"123"[0:2]
+
+b"123"[:1]
+
+b"123"[1:]
+
+# Idiom for copying sequence
+b"123"[:]
+
+b"123"[:-1]
+
+# Weird cases
+b"123"[0:0]
+b"123"[1:0]
+b"123"[1:1]
+b"123"[-1:-1]
+b"123"[-3:]
+b"123"[-3:3]
+b"123"[0:]
+b"123"[:0]
+b"123"[:-3]
+b"123"[:-4]
+# No IndexError!
+b""[1:1]
+b""[-1:-1]

unix-cpy/Makefile

@@ -43,6 +43,7 @@ PY_O = \
 	objnone.o \
 	objrange.o \
 	objset.o \
+	objslice.o \
 	objstr.o \
 	objtuple.o \
 	objtype.o \

unix/Makefile

@@ -50,6 +50,7 @@ PY_O = \
 	objnone.o \
 	objrange.o \
 	objset.o \
+	objslice.o \
 	objstr.o \
 	objtuple.o \
 	objtype.o \