+static void tokenize(char* raw, struct command_t *cmd)
+{
+ memset(cmd, '\0', sizeof(struct command_t));
+
+ /* Bound prefix on the %s would be needed to prevent buffer
+ * overflow. but we shortstop this more simply by making each
+ * raw-input buffer as long as the enrire inout buffer. */
+ sscanf(raw, "%s%s", cmd->raw1, cmd->raw2);
+
+ /* (ESR) In oldstyle mode, simulate the uppercasing and truncating
+ * effect on raw tokens of packing them into sixbit characters, 5
+ * to a 32-bit word. This is something the FORTRAN version did
+ * becuse archaic FORTRAN had no string types. Don Wood's
+ * mechanical translation of 2.5 to C retained the packing and
+ * thus this misfeature.
+ *
+ * It's philosophically questionable whether this is the right
+ * thing to do even in oldstyle mode. On one hand, the text
+ * mangling was not authorial intent, but a result of limitations
+ * in their tools. On the other, not simulating this misbehavior
+ * goes against the goal of making oldstyle as accurate as
+ * possible an emulation of the original UI.
+ */
+ if (settings.oldstyle) {
+ cmd->raw1[TOKLEN + TOKLEN] = cmd->raw2[TOKLEN + TOKLEN] = '\0';
+ for (size_t i = 0; i < strlen(cmd->raw1); i++)
+ cmd->raw1[i] = toupper(cmd->raw1[i]);
+ for (size_t i = 0; i < strlen(cmd->raw2); i++)
+ cmd->raw2[i] = toupper(cmd->raw2[i]);
+ }
+
+ /* populate command with parsed vocab metadata */
+ get_vocab_metadata(cmd->raw1, &(cmd->id1), &(cmd->type1));
+ get_vocab_metadata(cmd->raw2, &(cmd->id2), &(cmd->type2));
+}
+
+bool get_command_input(struct command_t *command)
+/* Get user input on stdin, parse and map to command */
+{
+ char inputbuf[LINESIZE];
+ char* input;
+
+ for (;;) {
+ input = get_input();
+ if (input == NULL)
+ return false;
+ if (word_count(input) > 2) {
+ rspeak(TWO_WORDS);
+ free(input);
+ continue;
+ }
+ if (strcmp(input, "") != 0)
+ break;
+ free(input);
+ }
+
+ strncpy(inputbuf, input, LINESIZE - 1);
+ free(input);
+
+ tokenize(inputbuf, command);
+
+ return true;
+}
+