diff options
Diffstat (limited to 'Python/compile.c')
| -rw-r--r-- | Python/compile.c | 236 |
1 files changed, 176 insertions, 60 deletions
diff --git a/Python/compile.c b/Python/compile.c index 2c4f7d7b41..92fea3d13c 100644 --- a/Python/compile.c +++ b/Python/compile.c @@ -171,7 +171,6 @@ static int compiler_addop(struct compiler *, int); static int compiler_addop_o(struct compiler *, int, PyObject *, PyObject *); static int compiler_addop_i(struct compiler *, int, Py_ssize_t); static int compiler_addop_j(struct compiler *, int, basicblock *, int); -static basicblock *compiler_use_new_block(struct compiler *); static int compiler_error(struct compiler *, const char *); static int compiler_nameop(struct compiler *, identifier, expr_context_ty); @@ -196,7 +195,7 @@ static int expr_constant(struct compiler *, expr_ty); static int compiler_with(struct compiler *, stmt_ty, int); static int compiler_async_with(struct compiler *, stmt_ty, int); static int compiler_async_for(struct compiler *, stmt_ty); -static int compiler_call_helper(struct compiler *c, Py_ssize_t n, +static int compiler_call_helper(struct compiler *c, int n, asdl_seq *args, asdl_seq *keywords); static int compiler_try_except(struct compiler *, stmt_ty); @@ -477,9 +476,9 @@ compiler_unit_check(struct compiler_unit *u) { basicblock *block; for (block = u->u_blocks; block != NULL; block = block->b_list) { - assert((void *)block != (void *)0xcbcbcbcb); - assert((void *)block != (void *)0xfbfbfbfb); - assert((void *)block != (void *)0xdbdbdbdb); + assert((Py_uintptr_t)block != 0xcbcbcbcbU); + assert((Py_uintptr_t)block != 0xfbfbfbfbU); + assert((Py_uintptr_t)block != 0xdbdbdbdbU); if (block->b_instr != NULL) { assert(block->b_ialloc > 0); assert(block->b_iused > 0); @@ -523,6 +522,7 @@ compiler_enter_scope(struct compiler *c, identifier name, int scope_type, void *key, int lineno) { struct compiler_unit *u; + basicblock *block; u = (struct compiler_unit *)PyObject_Malloc(sizeof( struct compiler_unit)); @@ -620,8 +620,11 @@ compiler_enter_scope(struct compiler *c, identifier name, c->u = u; c->c_nestlevel++; - if (compiler_use_new_block(c) == NULL) + + block = compiler_new_block(c); + if (block == NULL) return 0; + c->u->u_curblock = block; if (u->u_scope_type != COMPILER_SCOPE_MODULE) { if (!compiler_set_qualname(c)) @@ -731,6 +734,7 @@ compiler_set_qualname(struct compiler *c) return 1; } + /* Allocate a new block and return a pointer to it. Returns NULL on error. */ @@ -755,16 +759,6 @@ compiler_new_block(struct compiler *c) } static basicblock * -compiler_use_new_block(struct compiler *c) -{ - basicblock *block = compiler_new_block(c); - if (block == NULL) - return NULL; - c->u->u_curblock = block; - return block; -} - -static basicblock * compiler_next_block(struct compiler *c) { basicblock *block = compiler_new_block(c); @@ -1066,6 +1060,10 @@ PyCompile_OpcodeStackEffect(int opcode, int oparg) return 1; case GET_YIELD_FROM_ITER: return 0; + case FORMAT_VALUE: + /* If there's a fmt_spec on the stack, we go from 2->1, + else 1->1. */ + return (oparg & FVS_MASK) == FVS_HAVE_SPEC ? -1 : 0; default: return PY_INVALID_STACK_EFFECT; } @@ -1165,10 +1163,14 @@ compiler_addop_i(struct compiler *c, int opcode, Py_ssize_t oparg) struct instr *i; int off; - /* Integer arguments are limit to 16-bit. There is an extension for 32-bit - integer arguments. */ - assert((-2147483647-1) <= oparg); - assert(oparg <= 2147483647); + /* oparg value is unsigned, but a signed C int is usually used to store + it in the C code (like Python/ceval.c). + + Limit to 32-bit signed C int (rather than INT_MAX) for portability. + + The argument of a concrete bytecode instruction is limited to 16-bit. + EXTENDED_ARG is used for 32-bit arguments. */ + assert(0 <= oparg && oparg <= 2147483647); off = compiler_next_instr(c, c->u->u_curblock); if (off < 0) @@ -1203,22 +1205,12 @@ compiler_addop_j(struct compiler *c, int opcode, basicblock *b, int absolute) return 1; } -/* The distinction between NEW_BLOCK and NEXT_BLOCK is subtle. (I'd - like to find better names.) NEW_BLOCK() creates a new block and sets - it as the current block. NEXT_BLOCK() also creates an implicit jump - from the current block to the new block. -*/ +/* NEXT_BLOCK() creates an implicit jump from the current block + to the new block. -/* The returns inside these macros make it impossible to decref objects - created in the local function. Local objects should use the arena. + The returns inside this macro make it impossible to decref objects + created in the local function. Local objects should use the arena. */ - - -#define NEW_BLOCK(C) { \ - if (compiler_use_new_block((C)) == NULL) \ - return 0; \ -} - #define NEXT_BLOCK(C) { \ if (compiler_next_block((C)) == NULL) \ return 0; \ @@ -1309,7 +1301,11 @@ compiler_isdocstring(stmt_ty s) { if (s->kind != Expr_kind) return 0; - return s->v.Expr.value->kind == Str_kind; + if (s->v.Expr.value->kind == Str_kind) + return 1; + if (s->v.Expr.value->kind == Constant_kind) + return PyUnicode_CheckExact(s->v.Expr.value->v.Constant.value); + return 0; } /* Compile a sequence of statements, checking for a docstring. */ @@ -1683,8 +1679,12 @@ compiler_function(struct compiler *c, stmt_ty s, int is_async) st = (stmt_ty)asdl_seq_GET(body, 0); docstring = compiler_isdocstring(st); - if (docstring && c->c_optimize < 2) - first_const = st->v.Expr.value->v.Str.s; + if (docstring && c->c_optimize < 2) { + if (st->v.Expr.value->kind == Constant_kind) + first_const = st->v.Expr.value->v.Constant.value; + else + first_const = st->v.Expr.value->v.Str.s; + } if (compiler_add_o(c, c->u->u_consts, first_const) < 0) { compiler_exit_scope(c); return 0; @@ -2595,6 +2595,35 @@ compiler_assert(struct compiler *c, stmt_ty s) } static int +compiler_visit_stmt_expr(struct compiler *c, expr_ty value) +{ + if (c->c_interactive && c->c_nestlevel <= 1) { + VISIT(c, expr, value); + ADDOP(c, PRINT_EXPR); + return 1; + } + + switch (value->kind) + { + case Str_kind: + case Num_kind: + case Ellipsis_kind: + case Bytes_kind: + case NameConstant_kind: + case Constant_kind: + /* ignore constant statement */ + return 1; + + default: + break; + } + + VISIT(c, expr, value); + ADDOP(c, POP_TOP); + return 1; +} + +static int compiler_visit_stmt(struct compiler *c, stmt_ty s) { Py_ssize_t i, n; @@ -2664,16 +2693,7 @@ compiler_visit_stmt(struct compiler *c, stmt_ty s) case Nonlocal_kind: break; case Expr_kind: - if (c->c_interactive && c->c_nestlevel <= 1) { - VISIT(c, expr, s->v.Expr.value); - ADDOP(c, PRINT_EXPR); - } - else if (s->v.Expr.value->kind != Str_kind && - s->v.Expr.value->kind != Num_kind) { - VISIT(c, expr, s->v.Expr.value); - ADDOP(c, POP_TOP); - } - break; + return compiler_visit_stmt_expr(c, s->v.Expr.value); case Pass_kind: break; case Break_kind: @@ -3077,7 +3097,8 @@ compiler_set(struct compiler *c, expr_ty e) static int compiler_dict(struct compiler *c, expr_ty e) { - Py_ssize_t i, n, containers, elements; + Py_ssize_t i, n, elements; + int containers; int is_unpacking = 0; n = asdl_seq_LEN(e->v.Dict.values); containers = 0; @@ -3169,15 +3190,91 @@ compiler_call(struct compiler *c, expr_ty e) e->v.Call.keywords); } +static int +compiler_joined_str(struct compiler *c, expr_ty e) +{ + /* Concatenate parts of a string using ''.join(parts). There are + probably better ways of doing this. + + This is used for constructs like "'x=' f'{42}'", which have to + be evaluated at compile time. */ + + static PyObject *empty_string; + static PyObject *join_string; + + if (!empty_string) { + empty_string = PyUnicode_FromString(""); + if (!empty_string) + return 0; + } + if (!join_string) { + join_string = PyUnicode_FromString("join"); + if (!join_string) + return 0; + } + + ADDOP_O(c, LOAD_CONST, empty_string, consts); + ADDOP_NAME(c, LOAD_ATTR, join_string, names); + VISIT_SEQ(c, expr, e->v.JoinedStr.values); + ADDOP_I(c, BUILD_LIST, asdl_seq_LEN(e->v.JoinedStr.values)); + ADDOP_I(c, CALL_FUNCTION, 1); + return 1; +} + +/* Used to implement f-strings. Format a single value. */ +static int +compiler_formatted_value(struct compiler *c, expr_ty e) +{ + /* Our oparg encodes 2 pieces of information: the conversion + character, and whether or not a format_spec was provided. + + Convert the conversion char to 2 bits: + None: 000 0x0 FVC_NONE + !s : 001 0x1 FVC_STR + !r : 010 0x2 FVC_REPR + !a : 011 0x3 FVC_ASCII + + next bit is whether or not we have a format spec: + yes : 100 0x4 + no : 000 0x0 + */ + + int oparg; + + /* Evaluate the expression to be formatted. */ + VISIT(c, expr, e->v.FormattedValue.value); + + switch (e->v.FormattedValue.conversion) { + case 's': oparg = FVC_STR; break; + case 'r': oparg = FVC_REPR; break; + case 'a': oparg = FVC_ASCII; break; + case -1: oparg = FVC_NONE; break; + default: + PyErr_SetString(PyExc_SystemError, + "Unrecognized conversion character"); + return 0; + } + if (e->v.FormattedValue.format_spec) { + /* Evaluate the format spec, and update our opcode arg. */ + VISIT(c, expr, e->v.FormattedValue.format_spec); + oparg |= FVS_HAVE_SPEC; + } + + /* And push our opcode and oparg */ + ADDOP_I(c, FORMAT_VALUE, oparg); + return 1; +} + /* shared code between compiler_call and compiler_class */ static int compiler_call_helper(struct compiler *c, - Py_ssize_t n, /* Args already pushed */ + int n, /* Args already pushed */ asdl_seq *args, asdl_seq *keywords) { int code = 0; - Py_ssize_t nelts, i, nseen, nkw; + Py_ssize_t nelts, i, nseen; + int nkw; /* the number of tuples and dictionaries on the stack */ Py_ssize_t nsubargs = 0, nsubkwargs = 0; @@ -3545,6 +3642,8 @@ expr_constant(struct compiler *c, expr_ty e) switch (e->kind) { case Ellipsis_kind: return 1; + case Constant_kind: + return PyObject_IsTrue(e->v.Constant.value); case Num_kind: return PyObject_IsTrue(e->v.Num.n); case Str_kind: @@ -3588,9 +3687,9 @@ expr_constant(struct compiler *c, expr_ty e) BLOCK finally: if an exception was raised: - exc = copy of (exception, instance, traceback) + exc = copy of (exception, instance, traceback) else: - exc = (None, None, None) + exc = (None, None, None) if not (await exit(*exc)): raise */ @@ -3832,12 +3931,19 @@ compiler_visit_expr(struct compiler *c, expr_ty e) return compiler_compare(c, e); case Call_kind: return compiler_call(c, e); + case Constant_kind: + ADDOP_O(c, LOAD_CONST, e->v.Constant.value, consts); + break; case Num_kind: ADDOP_O(c, LOAD_CONST, e->v.Num.n, consts); break; case Str_kind: ADDOP_O(c, LOAD_CONST, e->v.Str.s, consts); break; + case JoinedStr_kind: + return compiler_joined_str(c, e); + case FormattedValue_kind: + return compiler_formatted_value(c, e); case Bytes_kind: ADDOP_O(c, LOAD_CONST, e->v.Bytes.s, consts); break; @@ -4329,7 +4435,6 @@ assemble_lnotab(struct assembler *a, struct instr *i) d_lineno = i->i_lineno - a->a_lineno; assert(d_bytecode >= 0); - assert(d_lineno >= 0); if(d_bytecode == 0 && d_lineno == 0) return 1; @@ -4359,9 +4464,21 @@ assemble_lnotab(struct assembler *a, struct instr *i) d_bytecode -= ncodes * 255; a->a_lnotab_off += ncodes * 2; } - assert(d_bytecode <= 255); - if (d_lineno > 255) { - int j, nbytes, ncodes = d_lineno / 255; + assert(0 <= d_bytecode && d_bytecode <= 255); + + if (d_lineno < -128 || 127 < d_lineno) { + int j, nbytes, ncodes, k; + if (d_lineno < 0) { + k = -128; + /* use division on positive numbers */ + ncodes = (-d_lineno) / 128; + } + else { + k = 127; + ncodes = d_lineno / 127; + } + d_lineno -= ncodes * k; + assert(ncodes >= 1); nbytes = a->a_lnotab_off + 2 * ncodes; len = PyBytes_GET_SIZE(a->a_lnotab); if (nbytes >= len) { @@ -4379,15 +4496,15 @@ assemble_lnotab(struct assembler *a, struct instr *i) lnotab = (unsigned char *) PyBytes_AS_STRING(a->a_lnotab) + a->a_lnotab_off; *lnotab++ = d_bytecode; - *lnotab++ = 255; + *lnotab++ = k; d_bytecode = 0; for (j = 1; j < ncodes; j++) { *lnotab++ = 0; - *lnotab++ = 255; + *lnotab++ = k; } - d_lineno -= ncodes * 255; a->a_lnotab_off += ncodes * 2; } + assert(-128 <= d_lineno && d_lineno <= 127); len = PyBytes_GET_SIZE(a->a_lnotab); if (a->a_lnotab_off + 2 >= len) { @@ -4745,4 +4862,3 @@ PyAST_Compile(mod_ty mod, const char *filename, PyCompilerFlags *flags, { return PyAST_CompileEx(mod, filename, flags, -1, arena); } - |