STATIC mp_obj_t socket_recvfrom(size_t n_args, const mp_obj_t *args) { mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]); int sz = MP_OBJ_SMALL_INT_VALUE(args[1]); int flags = 0; if (n_args > 2) { flags = MP_OBJ_SMALL_INT_VALUE(args[2]); } struct sockaddr_storage addr; socklen_t addr_len = sizeof(addr); byte *buf = m_new(byte, sz); int out_sz = recvfrom(self->fd, buf, sz, flags, (struct sockaddr*)&addr, &addr_len); RAISE_ERRNO(out_sz, errno); mp_obj_t buf_o = mp_obj_new_str_of_type(&mp_type_bytes, buf, out_sz); m_del(char, buf, sz); mp_obj_tuple_t *t = MP_OBJ_TO_PTR(mp_obj_new_tuple(2, NULL)); t->items[0] = buf_o; t->items[1] = mp_obj_from_sockaddr((struct sockaddr*)&addr, addr_len); return MP_OBJ_FROM_PTR(t); }
static mp_obj_t translate(mp_obj_t words) { mp_uint_t len, outlen; const char *txt = mp_obj_str_get_data(words, &len); // Reciter truncates *output* at about 120 characters. // So to avoid that we must disallow any input that will exceed that. if (len > 80) { nlr_raise(mp_obj_new_exception_msg(&mp_type_ValueError, "text too long.")); } reciter_memory *mem = m_new(reciter_memory, 1); MP_STATE_PORT(speech_data) = mem; for (mp_uint_t i = 0; i < len; i++) { mem->input[i] = txt[i]; } mem->input[len] = '['; if (!TextToPhonemes(mem)) { MP_STATE_PORT(speech_data) = NULL; nlr_raise(mp_obj_new_exception_msg(&mp_type_ValueError, "could not parse input.")); } for (outlen = 0; outlen < 255; outlen++) { if (mem->input[outlen] == 155) { break; } } mp_obj_t res = mp_obj_new_str_of_type(&mp_type_str, (byte *)mem->input, outlen); // Prevent input becoming invisible to GC due to tail-call optimisation. MP_STATE_PORT(speech_data) = NULL; return res; }MP_DEFINE_CONST_FUN_OBJ_1(translate_obj, translate);
STATIC void push_result_token(parser_t *parser, const rule_t *rule) { mp_parse_node_t pn; mp_lexer_t *lex = parser->lexer; if (lex->tok_kind == MP_TOKEN_NAME) { qstr id = qstr_from_strn(lex->vstr.buf, lex->vstr.len); #if MICROPY_COMP_CONST // if name is a standalone identifier, look it up in the table of dynamic constants mp_map_elem_t *elem; if (rule->rule_id == RULE_atom && (elem = mp_map_lookup(&parser->consts, MP_OBJ_NEW_QSTR(id), MP_MAP_LOOKUP)) != NULL) { if (MP_OBJ_IS_SMALL_INT(elem->value)) { pn = mp_parse_node_new_small_int(MP_OBJ_SMALL_INT_VALUE(elem->value)); } else { pn = make_node_const_object(parser, lex->tok_line, elem->value); } } else { pn = mp_parse_node_new_leaf(MP_PARSE_NODE_ID, id); } #else (void)rule; pn = mp_parse_node_new_leaf(MP_PARSE_NODE_ID, id); #endif } else if (lex->tok_kind == MP_TOKEN_INTEGER) { mp_obj_t o = mp_parse_num_integer(lex->vstr.buf, lex->vstr.len, 0, lex); if (MP_OBJ_IS_SMALL_INT(o)) { pn = mp_parse_node_new_small_int(MP_OBJ_SMALL_INT_VALUE(o)); } else { pn = make_node_const_object(parser, lex->tok_line, o); } } else if (lex->tok_kind == MP_TOKEN_FLOAT_OR_IMAG) { mp_obj_t o = mp_parse_num_decimal(lex->vstr.buf, lex->vstr.len, true, false, lex); pn = make_node_const_object(parser, lex->tok_line, o); } else if (lex->tok_kind == MP_TOKEN_STRING || lex->tok_kind == MP_TOKEN_BYTES) { // Don't automatically intern all strings/bytes. doc strings (which are usually large) // will be discarded by the compiler, and so we shouldn't intern them. qstr qst = MP_QSTR_NULL; if (lex->vstr.len <= MICROPY_ALLOC_PARSE_INTERN_STRING_LEN) { // intern short strings qst = qstr_from_strn(lex->vstr.buf, lex->vstr.len); } else { // check if this string is already interned qst = qstr_find_strn(lex->vstr.buf, lex->vstr.len); } if (qst != MP_QSTR_NULL) { // qstr exists, make a leaf node pn = mp_parse_node_new_leaf(lex->tok_kind == MP_TOKEN_STRING ? MP_PARSE_NODE_STRING : MP_PARSE_NODE_BYTES, qst); } else { // not interned, make a node holding a pointer to the string/bytes object mp_obj_t o = mp_obj_new_str_of_type( lex->tok_kind == MP_TOKEN_STRING ? &mp_type_str : &mp_type_bytes, (const byte*)lex->vstr.buf, lex->vstr.len); pn = make_node_const_object(parser, lex->tok_line, o); } } else { pn = mp_parse_node_new_leaf(MP_PARSE_NODE_TOKEN, lex->tok_kind); } push_result_node(parser, pn); }
// Version of mp_vfs_lookup_path that takes and returns uPy string objects. STATIC mp_vfs_mount_t *lookup_path(mp_obj_t path_in, mp_obj_t *path_out) { const char *path = mp_obj_str_get_str(path_in); const char *p_out; mp_vfs_mount_t *vfs = mp_vfs_lookup_path(path, &p_out); if (vfs != MP_VFS_NONE && vfs != MP_VFS_ROOT) { *path_out = mp_obj_new_str_of_type(mp_obj_get_type(path_in), (const byte*)p_out, strlen(p_out)); } return vfs; }
STATIC mp_obj_t stream_readall(mp_obj_t self_in) { struct _mp_obj_base_t *o = (struct _mp_obj_base_t *)self_in; if (o->type->stream_p == NULL || o->type->stream_p->read == NULL) { // CPython: io.UnsupportedOperation, OSError subclass nlr_raise(mp_obj_new_exception_msg(&mp_type_OSError, "Operation not supported")); } int total_size = 0; vstr_t *vstr = vstr_new_size(DEFAULT_BUFFER_SIZE); char *buf = vstr_str(vstr); char *p = buf; int error; int current_read = DEFAULT_BUFFER_SIZE; while (true) { mp_int_t out_sz = o->type->stream_p->read(self_in, p, current_read, &error); if (out_sz == -1) { if (is_nonblocking_error(error)) { // With non-blocking streams, we read as much as we can. // If we read nothing, return None, just like read(). // Otherwise, return data read so far. if (total_size == 0) { return mp_const_none; } break; } nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_OSError, "[Errno %d]", error)); } if (out_sz == 0) { break; } total_size += out_sz; if (out_sz < current_read) { current_read -= out_sz; p += out_sz; } else { current_read = DEFAULT_BUFFER_SIZE; p = vstr_extend(vstr, current_read); if (p == NULL) { // TODO nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_OSError/*&mp_type_RuntimeError*/, "Out of memory")); } } } mp_obj_t s = mp_obj_new_str_of_type(STREAM_CONTENT_TYPE(o->type->stream_p), (byte*)vstr->buf, total_size); vstr_free(vstr); return s; }
// Unbuffered, inefficient implementation of readline() for raw I/O files. STATIC mp_obj_t stream_unbuffered_readline(uint n_args, const mp_obj_t *args) { struct _mp_obj_base_t *o = (struct _mp_obj_base_t *)args[0]; if (o->type->stream_p == NULL || o->type->stream_p->read == NULL) { // CPython: io.UnsupportedOperation, OSError subclass nlr_raise(mp_obj_new_exception_msg(&mp_type_OSError, "Operation not supported")); } mp_int_t max_size = -1; if (n_args > 1) { max_size = MP_OBJ_SMALL_INT_VALUE(args[1]); } vstr_t *vstr; if (max_size != -1) { vstr = vstr_new_size(max_size); } else { vstr = vstr_new(); } int error; while (max_size == -1 || max_size-- != 0) { char *p = vstr_add_len(vstr, 1); if (p == NULL) { nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_MemoryError, "out of memory")); } mp_int_t out_sz = o->type->stream_p->read(o, p, 1, &error); if (out_sz == -1) { nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_OSError, "[Errno %d]", error)); } if (out_sz == 0) { // Back out previously added byte // TODO: This is a bit hacky, does it supported by vstr API contract? // Consider, what's better - read a char and get OutOfMemory (so read // char is lost), or allocate first as we do. vstr_add_len(vstr, -1); break; } if (*p == '\n') { break; } } // TODO need a string creation API that doesn't copy the given data mp_obj_t ret = mp_obj_new_str_of_type(STREAM_CONTENT_TYPE(o->type->stream_p), (byte*)vstr->buf, vstr->len); vstr_free(vstr); return ret; }
// Note: besides flag param, this differs from read() in that // this does not swallow blocking errors (EAGAIN, EWOULDBLOCK) - // these would be thrown as exceptions. STATIC mp_obj_t socket_recv(size_t n_args, const mp_obj_t *args) { mp_obj_socket_t *self = MP_OBJ_TO_PTR(args[0]); int sz = MP_OBJ_SMALL_INT_VALUE(args[1]); int flags = 0; if (n_args > 2) { flags = MP_OBJ_SMALL_INT_VALUE(args[2]); } byte *buf = m_new(byte, sz); int out_sz = recv(self->fd, buf, sz, flags); RAISE_ERRNO(out_sz, errno); mp_obj_t ret = mp_obj_new_str_of_type(&mp_type_bytes, buf, out_sz); m_del(char, buf, sz); return ret; }
STATIC mp_obj_t str_subscr(mp_obj_t self_in, mp_obj_t index, mp_obj_t value) { mp_obj_type_t *type = mp_obj_get_type(self_in); assert(type == &mp_type_str); GET_STR_DATA_LEN(self_in, self_data, self_len); if (value == MP_OBJ_SENTINEL) { // load #if MICROPY_PY_BUILTINS_SLICE if (MP_OBJ_IS_TYPE(index, &mp_type_slice)) { mp_obj_t ostart, ostop, ostep; mp_obj_slice_get(index, &ostart, &ostop, &ostep); if (ostep != mp_const_none && ostep != MP_OBJ_NEW_SMALL_INT(1)) { nlr_raise(mp_obj_new_exception_msg(&mp_type_NotImplementedError, "only slices with step=1 (aka None) are supported")); } const byte *pstart, *pstop; if (ostart != mp_const_none) { pstart = str_index_to_ptr(type, self_data, self_len, ostart, true); } else { pstart = self_data; } if (ostop != mp_const_none) { // pstop will point just after the stop character. This depends on // the \0 at the end of the string. pstop = str_index_to_ptr(type, self_data, self_len, ostop, true); } else { pstop = self_data + self_len; } if (pstop < pstart) { return MP_OBJ_NEW_QSTR(MP_QSTR_); } return mp_obj_new_str_of_type(type, (const byte *)pstart, pstop - pstart); } #endif const byte *s = str_index_to_ptr(type, self_data, self_len, index, false); int len = 1; if (UTF8_IS_NONASCII(*s)) { // Count the number of 1 bits (after the first) for (char mask = 0x40; *s & mask; mask >>= 1) { ++len; } } return mp_obj_new_str((const char*)s, len, true); // This will create a one-character string } else { return MP_OBJ_NULL; // op not supported
STATIC mp_obj_t str_make_new(mp_obj_t type_in, mp_uint_t n_args, mp_uint_t n_kw, const mp_obj_t *args) { #if MICROPY_CPYTHON_COMPAT if (n_kw != 0) { mp_arg_error_unimpl_kw(); } #endif switch (n_args) { case 0: return MP_OBJ_NEW_QSTR(MP_QSTR_); case 1: { vstr_t *vstr = vstr_new(); mp_obj_print_helper((void (*)(void*, const char*, ...))vstr_printf, vstr, args[0], PRINT_STR); mp_obj_t s = mp_obj_new_str(vstr->buf, vstr->len, false); vstr_free(vstr); return s; } case 2: case 3: { // TODO: validate 2nd/3rd args if (MP_OBJ_IS_TYPE(args[0], &mp_type_bytes)) { GET_STR_DATA_LEN(args[0], str_data, str_len); GET_STR_HASH(args[0], str_hash); mp_obj_str_t *o = mp_obj_new_str_of_type(&mp_type_str, NULL, str_len); o->data = str_data; o->hash = str_hash; return o; } else { mp_buffer_info_t bufinfo; mp_get_buffer_raise(args[0], &bufinfo, MP_BUFFER_READ); return mp_obj_new_str(bufinfo.buf, bufinfo.len, false); } } default: nlr_raise(mp_obj_new_exception_msg(&mp_type_TypeError, "str takes at most 3 arguments")); } }
STATIC mp_obj_t stream_read(uint n_args, const mp_obj_t *args) { struct _mp_obj_base_t *o = (struct _mp_obj_base_t *)args[0]; if (o->type->stream_p == NULL || o->type->stream_p->read == NULL) { // CPython: io.UnsupportedOperation, OSError subclass nlr_raise(mp_obj_new_exception_msg(&mp_type_OSError, "Operation not supported")); } mp_int_t sz; if (n_args == 1 || ((sz = mp_obj_get_int(args[1])) == -1)) { return stream_readall(args[0]); } #if MICROPY_PY_BUILTINS_STR_UNICODE if (!o->type->stream_p->is_bytes) { mp_not_implemented("Reading from unicode text streams by character count"); } #endif byte *buf = m_new(byte, sz); int error; mp_int_t out_sz = o->type->stream_p->read(o, buf, sz, &error); if (out_sz == -1) { if (is_nonblocking_error(error)) { // https://docs.python.org/3.4/library/io.html#io.RawIOBase.read // "If the object is in non-blocking mode and no bytes are available, // None is returned." // This is actually very weird, as naive truth check will treat // this as EOF. return mp_const_none; } nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_OSError, "[Errno %d]", error)); } else { mp_obj_t s = mp_obj_new_str_of_type(STREAM_CONTENT_TYPE(o->type->stream_p), buf, out_sz); // will reallocate to use exact size m_free(buf, sz); return s; } }
STATIC mp_obj_t stream_read(uint n_args, const mp_obj_t *args) { struct _mp_obj_base_t *o = (struct _mp_obj_base_t *)args[0]; if (o->type->stream_p == NULL || o->type->stream_p->read == NULL) { // CPython: io.UnsupportedOperation, OSError subclass nlr_raise(mp_obj_new_exception_msg(&mp_type_OSError, "Operation not supported")); } // What to do if sz < -1? Python docs don't specify this case. // CPython does a readall, but here we silently let negatives through, // and they will cause a MemoryError. mp_int_t sz; if (n_args == 1 || ((sz = mp_obj_get_int(args[1])) == -1)) { return stream_readall(args[0]); } #if MICROPY_PY_BUILTINS_STR_UNICODE if (o->type->stream_p->is_text) { // We need to read sz number of unicode characters. Because we don't have any // buffering, and because the stream API can only read bytes, we must read here // in units of bytes and must never over read. If we want sz chars, then reading // sz bytes will never over-read, so we follow this approach, in a loop to keep // reading until we have exactly enough chars. This will be 1 read for text // with ASCII-only chars, and about 2 reads for text with a couple of non-ASCII // chars. For text with lots of non-ASCII chars, it'll be pretty inefficient // in time and memory. vstr_t vstr; vstr_init(&vstr, sz); mp_uint_t more_bytes = sz; mp_uint_t last_buf_offset = 0; while (more_bytes > 0) { char *p = vstr_add_len(&vstr, more_bytes); if (p == NULL) { nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_MemoryError, "out of memory")); } int error; mp_uint_t out_sz = o->type->stream_p->read(o, p, more_bytes, &error); if (out_sz == MP_STREAM_ERROR) { vstr_cut_tail_bytes(&vstr, more_bytes); if (is_nonblocking_error(error)) { // With non-blocking streams, we read as much as we can. // If we read nothing, return None, just like read(). // Otherwise, return data read so far. // TODO what if we have read only half a non-ASCII char? if (vstr.len == 0) { vstr_clear(&vstr); return mp_const_none; } break; } nlr_raise(mp_obj_new_exception_arg1(&mp_type_OSError, MP_OBJ_NEW_SMALL_INT(error))); } if (out_sz < more_bytes) { // Finish reading. // TODO what if we have read only half a non-ASCII char? vstr_cut_tail_bytes(&vstr, more_bytes - out_sz); if (out_sz == 0) { break; } } // count chars from bytes just read for (mp_uint_t off = last_buf_offset;;) { byte b = vstr.buf[off]; int n; if (!UTF8_IS_NONASCII(b)) { // 1-byte ASCII char n = 1; } else if ((b & 0xe0) == 0xc0) { // 2-byte char n = 2; } else if ((b & 0xf0) == 0xe0) { // 3-byte char n = 3; } else if ((b & 0xf8) == 0xf0) { // 4-byte char n = 4; } else { // TODO n = 5; } if (off + n <= vstr.len) { // got a whole char in n bytes off += n; sz -= 1; last_buf_offset = off; if (off >= vstr.len) { more_bytes = sz; break; } } else { // didn't get a whole char, so work out how many extra bytes are needed for // this partial char, plus bytes for additional chars that we want more_bytes = (off + n - vstr.len) + (sz - 1); break; } } } mp_obj_t ret = mp_obj_new_str_of_type(&mp_type_str, (byte*)vstr.buf, vstr.len); vstr_clear(&vstr); return ret; } #endif byte *buf = m_new(byte, sz); int error; mp_uint_t out_sz = o->type->stream_p->read(o, buf, sz, &error); if (out_sz == MP_STREAM_ERROR) { if (is_nonblocking_error(error)) { // https://docs.python.org/3.4/library/io.html#io.RawIOBase.read // "If the object is in non-blocking mode and no bytes are available, // None is returned." // This is actually very weird, as naive truth check will treat // this as EOF. return mp_const_none; } nlr_raise(mp_obj_new_exception_arg1(&mp_type_OSError, MP_OBJ_NEW_SMALL_INT(error))); } else { mp_obj_t s = mp_obj_new_str_of_type(STREAM_CONTENT_TYPE(o->type->stream_p), buf, out_sz); // will reallocate to use exact size m_free(buf, sz); return s; } }
// Unbuffered, inefficient implementation of readline() for raw I/O files. STATIC mp_obj_t stream_unbuffered_readline(uint n_args, const mp_obj_t *args) { struct _mp_obj_base_t *o = (struct _mp_obj_base_t *)args[0]; if (o->type->stream_p == NULL || o->type->stream_p->read == NULL) { // CPython: io.UnsupportedOperation, OSError subclass nlr_raise(mp_obj_new_exception_msg(&mp_type_OSError, "Operation not supported")); } mp_int_t max_size = -1; if (n_args > 1) { max_size = MP_OBJ_SMALL_INT_VALUE(args[1]); } vstr_t *vstr; if (max_size != -1) { vstr = vstr_new_size(max_size); } else { vstr = vstr_new(); } int error; while (max_size == -1 || max_size-- != 0) { char *p = vstr_add_len(vstr, 1); if (p == NULL) { nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_MemoryError, "out of memory")); } mp_uint_t out_sz = o->type->stream_p->read(o, p, 1, &error); if (out_sz == MP_STREAM_ERROR) { if (is_nonblocking_error(error)) { if (vstr->len == 1) { // We just incremented it, but otherwise we read nothing // and immediately got EAGAIN. This is case is not well // specified in // https://docs.python.org/3/library/io.html#io.IOBase.readline // unlike similar case for read(). But we follow the latter's // behavior - return None. vstr_free(vstr); return mp_const_none; } else { goto done; } } nlr_raise(mp_obj_new_exception_arg1(&mp_type_OSError, MP_OBJ_NEW_SMALL_INT(error))); } if (out_sz == 0) { done: // Back out previously added byte // Consider, what's better - read a char and get OutOfMemory (so read // char is lost), or allocate first as we do. vstr_cut_tail_bytes(vstr, 1); break; } if (*p == '\n') { break; } } // TODO need a string creation API that doesn't copy the given data mp_obj_t ret = mp_obj_new_str_of_type(STREAM_CONTENT_TYPE(o->type->stream_p), (byte*)vstr->buf, vstr->len); vstr_free(vstr); return ret; }
STATIC mp_obj_t stringio_getvalue(mp_obj_t self_in) { mp_obj_stringio_t *self = self_in; return mp_obj_new_str_of_type(STREAM_TO_CONTENT_TYPE(self), (byte*)self->vstr->buf, self->vstr->len); }
STATIC mp_obj_t stringio_getvalue(mp_obj_t self_in) { mp_obj_stringio_t *self = MP_OBJ_TO_PTR(self_in); check_stringio_is_open(self); // TODO: Try to avoid copying string return mp_obj_new_str_of_type(STREAM_TO_CONTENT_TYPE(self), (byte*)self->vstr->buf, self->vstr->len); }