5 The main task of the lexical scanner is to convert a stream of
6 characters into a stream of tokens. The tokens are then typically
7 used by a parser to extract the syntactic structure.
9 The stream of characters are assumed to be in memory identified by a
10 linked list of blocks, such as provided by the "[mdcode][]" literate
11 program extractor. A single token may never cross a block boundary.
18 The text is assumed to be UTF-8 though some matching assumes the
19 ASCII subset. If the text provided does not conform to UTF-8 an error
20 will be reported and some number of bytes will be skipped.
25 #include <unicode/uchar.h>
27 Tokens are returned by successive calls to the main interface
28 function: `token_next()` which has a `state` structure to keep track
29 of where it is up to. Each token carries not just a numeric
30 identifier but also the code block, the line and character within that
31 block, and the actual start and length using the `struct text` from
37 struct code_node *node;
48 ###### exported functions
49 struct token token_next(struct token_state *state);
52 struct token token_next(struct token_state *state)
63 The `line` and `col` offsets are useful for reporting errors.
64 The `txt` provides the content when that is important.
66 ### Token types and configuration ##
68 The scanner is not completely general, yet not completely specified.
69 There are a fixed set of token types, though particular tokens within
70 those types can be distinguish via configuration.
72 Most token types may be explicitly ignored, as typically comments
73 would be. The exact consequence of ignoring each token type varies
78 int ignored; // bit set of ignored tokens.
79 ## token config parameters
83 struct token_config *conf;
85 ###### token_next init
86 int ignored = state->conf->ignored;
89 The different tokens are numbers, words, marks, strings, comments,
90 newlines, EOF, and indents, each of which is examined in detail below.
92 There are various cases where no token can be found in part of the
93 input. All of these will be reported as a `TK_error` token.
95 It is possible to declare a number of strings which form distinct
96 tokens (rather than being grouped as e.g. 'word'). These are given
97 token numbers from `TK_reserved` upwards.
108 Numbers are the messiest tokens to parse, primarily because they can
109 contain characters that also have meaning outside of numbers and,
110 particularly, immediately after numbers.
112 The obvious example is the '`-`' sign. It can come inside a number for
113 a negative exponent, or after a number as a subtraction operator. To
114 be sure we have parsed as best as possible we need to only allow the
115 '`-`' inside a number if it is after an exponent character. This can be
116 `e` or `p` (for hex exponents), but `e` can also be a hexadecimal
117 digit, so we don't allow '`-`' after just any `e`.
119 To make matters worse, our language designer has decided to experiment
120 with allowing commas to be used as the decimal indicator, and spaces
121 to be used to separate groups of digits in large numbers. Both of
122 these can reasonably be restricted to appear between two digits, so we
123 have to add that condition to our tests.
125 So we cannot just treat numbers as starting with a digit and being
126 followed by some set of characters. We need more structure than that.
130 - Numbers must start with a digit.
131 - If the first digit is zero, the next character must be a base
132 signifier (one of `xob`) or a decimal marker (`.` or `,`).
133 In the first case the first `p` or `P` may be followed by a sign.
134 - If the number doesn't start with `0` followed by one of `xob`, the
135 first `e` may be followed by a sign.
136 - Any digit or hex digit may be followed by a space or underscore
137 providing that the subsequence character is also a (hex) digit.
138 This rule will require an extra level of 'unget' to be
139 supported when handling characters.
140 - Otherwise any digits or ASCII letters are allowed. We do not at
141 this point check that the digits given are permitted by the base.
142 That will happen when the token is converted to a number.
144 To allow easy configuration, the various non alphanumeric characters
145 are only permitted if they are listed in a configuration parameter.
147 ###### token config parameters
150 Note that numbers may not start with a period, so `.75` is not a
151 number. This is not the norm, but is not unheard of. Excluding these
152 numbers simplifies the rule at very little cost.
157 If TK_number is ignored, digits will result in an error unless they
158 are declared to be a start character for words.
166 if (iswdigit(ch) && !(ignored & (1<<TK_number))) {
167 int prev_special = 0;
169 int decimal_mark = 0;
171 wchar_t ch2 = get_char(state);
172 if (strchr("xobXOB", ch2) != NULL)
180 if (ch == 'e' || ch == 'E')
184 if (ch == 'p' || ch == 'P')
188 save_unget_state(state);
189 ch = get_char(state);
194 if (ch == '+' || ch == '-') {
199 if (ch == '.' || ch == ',') {
205 /* Don't allow that special char,
208 restore_unget_state(state);
211 if (strchr(state->conf->number_chars, ch)) {
215 /* non-number char */
218 /* We seem to have a "number" token */
220 close_token(state, &tk);
226 Words start with a "start" character followed by the longest
227 sequence of "continue" characters. The Unicode ID_START and
228 ID_CONTINUE sets are always permitted, but other ASCII characters
229 can be added to these sets.
231 ###### token config parameters
235 ###### internal functions
236 static int is_word_start(wchar_t ch, struct token_config *conf)
238 return iswalpha(ch) ||
239 strchr(conf->word_start, ch) != NULL ||
240 u_hasBinaryProperty(ch, UCHAR_ID_START);
243 static int is_word_continue(wchar_t ch, struct token_config *conf)
245 return iswalnum(ch) ||
246 strchr(conf->word_cont, ch) != NULL ||
247 u_hasBinaryProperty(ch, UCHAR_ID_CONTINUE);
250 Words can be either known or unknown. Known words are referred to as
251 "reserved words" and get a unique token number. Unknown words are
252 "identifiers" and are syntactically a single token.
257 A list of known words must be provided. This list is shared with the
258 "marks" which are described next. The list must be lexically sorted
259 and the length of the list must be given (`known_count`).
260 Tokens matching these known words are reported as the index of the
261 list added to `TK_reserved`.
263 If identifiers are ignored, then any word which is not listed as a
264 known word results in an error.
266 ###### token config parameters
267 const char **words_marks;
272 if (is_word_start(ch, state->conf)) {
274 /* A word: identifier or reserved */
276 ch = get_char(state);
277 while (is_word_continue(ch, state->conf));
279 close_token(state, &tk);
281 if (ignored & (1<<TK_ident))
283 n = find_known(state->conf, tk.txt);
285 tk.num = TK_reserved + n;
291 Marks are generally one or more punctuation marks joined together. It
292 would be nice to use the term "symbol" for these, but that causes
293 confusion in a subsequent discussion of the grammar, which has terminal
294 symbols and non-terminal symbols which are conceptually quite
295 different. So strings of punctuation characters will be marks.
297 A "mark" consists of ASCII characters that are not white space, are not
298 "start" characters for words, and are not digits.
299 These will collectively be called mark characters.
301 ###### internal functions
302 static int is_mark(wchar_t ch, struct token_config *conf)
307 strchr(conf->word_start, ch) == NULL;
310 As with words, there can be known and unknown marks, though the rules
311 are slightly different.
313 Two marks do not need to be separated by a non-mark characters. This
314 is different from words which do need to be separated by at least one
315 non-continue character.
317 The scanner will normally prefer longer sequences of mark characters,
318 but will more strongly prefer known marks over unknown marks. So if
319 it finds a known mark where adding one more character does not result
320 in a known mark, it will return that first known mark.
322 If no known mark is found we will test against strings and comments
323 below before giving up and assuming an unknown mark.
325 If an unknown mark contains a quote character or a comment marker, and
326 that token is not being ignored, then we terminate the unknown mark
327 before that quote or comment. This ensures that an unknown mark
328 immediately before a string is handled correctly.
330 If the first character of a comment marker (i.e. '/') is a known mark,
331 the above rules would suggest that the start of a comment would be
332 parsed as that mark, which is not what is wanted. So the introductory
333 sequences for a comment ("//" and "/*") are treated as
334 partially-known. They prevent the leading "/" from being a mark by
335 itself, but do not actually constitute a stand-alone mark.
337 If `TK_mark` is ignored, then unknown marks are returned as errors.
342 Known marks are included in the same list as the list of known words.
346 while (is_mark(ch, state->conf)) {
349 close_token(state, &tk);
350 n = find_known(state->conf, tk.txt);
352 tk.num = TK_reserved + n;
353 else if (tk.num != TK_error) {
354 /* found a longest-known-mark, still need to
357 if (tk.txt.len == 2 && tk.txt.txt[0] == '/' &&
358 (ch == '/' || ch == '*')) {
359 /* Yes, this is a comment, not a '/' */
360 restore_unget_state(state);
365 close_token(state, &tk);
369 save_unget_state(state);
370 ch = get_char(state);
371 if (!(ignored && (1<<TK_string)) && is_quote(ch))
373 if (prev == '#' && n < 0)
374 /* '#' is not a known mark, so assume it is a comment */
376 if (prev == '/' && ch == '/' && tk.txt.len == 1 && n < 0) {
377 close_token(state, &tk);
378 restore_unget_state(state);
381 if (prev == '/' && ch == '*' && tk.txt.len == 1 && n < 0) {
382 close_token(state, &tk);
383 restore_unget_state(state);
388 if (tk.num != TK_error) {
389 close_token(state, &tk);
393 If we don't find a known mark, we will check for strings and comments
394 before assuming that we have an unknown mark
403 if (ignored & (1<<TK_mark))
412 Strings start with one of single quote, double quote, or back quote
413 and continue until a matching character on the same line. Any of
414 these characters can be included in the list of known marks and then
415 they will not be used for identifying strings.
417 Immediately following the close quote, one or two ASCII letters may
418 appear. These are somewhat like the arbitrary letters allowed in
419 "Numbers" above. They can be used by the language in various ways.
421 If 3 identical quote characters appear in a row and are
422 followed by a newline, then this forms a multi-line string which
423 continues until an identical triple quote appears on a line preceded
424 only by whitespace and followed immediately by 0-2 ASCII letters and a newline.
426 Multi-line strings may not extend beyond the end of the `code_node` in
429 Normal strings and multi-line strings are encoded as two different
436 ###### internal functions
437 static int is_quote(wchar_t ch)
439 return ch == '\'' || ch == '"' || ch == '`';
442 #### Multi-line strings
444 The multi-line string is checked for first. If they are being
445 ignored, we fall through and treat a triple quote as an empty string
446 followed by the start of a new string.
449 if (tk.txt.len == 3 &&
450 !(ignored & (1 << TK_multi_string)) &&
451 is_quote(tk.txt.txt[0]) &&
452 memcmp(tk.txt.txt, tk.txt.txt+1, 2) == 0 &&
453 is_newline(tk.txt.txt[3])) {
455 wchar_t first = tk.txt.txt[0];
458 while (!at_eon(state) && qseen < 3) {
459 ch = get_char(state);
460 if (is_newline(ch)) {
463 } else if (at_sol && ch == first) {
465 } else if (ch != ' ' && ch != '\t') {
471 /* Hit end of node - error.
472 * unget so the newline is seen,
473 * but return rest of string as an error.
477 close_token(state, &tk);
481 /* 2 letters are allowed */
482 ch = get_char(state);
484 ch = get_char(state);
486 ch = get_char(state);
487 /* Now we must have a newline, but we don't return it
490 close_token(state, &tk);
491 tk.num = TK_multi_string;
497 #### Single-line strings
499 The sequence of marks collected may be more than a single-line
500 string, so we reset to the start and collect characters until
501 we find a close quote or a newline.
503 If `TK_string` is ignored, then quote characters will appear as `TK_mark`s.
506 if (tk.txt.len && is_quote(tk.txt.txt[0]) &&
507 !(ignored & (1<<TK_string))) {
508 wchar_t first = tk.txt.txt[0];
509 reset_token(state, &tk);
510 ch = get_char(state);
512 while (!at_eon(state) && !is_newline(ch)) {
513 ch = get_char(state);
518 if (is_newline(ch)) {
523 while (!at_eon(state) && (ch = get_char(state)) &&
527 close_token(state, &tk);
533 Single line comments may start with '`//`' or '`#`' providing that these
534 are not known marks. They continue to the end of the line.
536 Block comments start with '`/*`' if this is not a known mark. They
537 continue to the first occurrence of '`*/`' and may not contain any
538 occurrence of '`/*`'.
540 Block comments can be wholly within one line or can continue over
541 multiple lines. The multi-line version should be followed immediately
542 by a newline. The Linux kernel contains over 285000 multi-line
543 comments are only 34 are followed by characters other than white space
544 (which should be removed) or a backslash (only needed in macros). So
545 it would not suffer from this rule.
547 These two comment types are reported as two separate token types, and
548 consequently can be ignored separately. When ignored a comment is
549 still parsed, but is discarded.
555 ###### internal functions
556 static int is_line_comment(struct text txt)
558 return (txt.len >= 1 && txt.txt[0] == '#') ||
559 (txt.len >= 2 && txt.txt[0] == '/' &&
563 static int is_block_comment(struct text txt)
565 return txt.len >= 2 && txt.txt[0] == '/' &&
569 #### Single line comments
571 A single-line comment continues up to, but not including the newline
576 if (is_line_comment(tk.txt)) {
577 while (!is_newline(ch) && !at_eon(state))
578 ch = get_char(state);
581 close_token(state, &tk);
582 tk.num = TK_line_comment;
583 if (ignored & (1 << TK_line_comment))
590 The token text collected so far could exceed the comment, so we need
593 If we find an embedded `/*` we reset to just before the '/' and report
594 an error. That way the next thing to be parsed will be the rest of
595 the comment. This requires a double unget, so we need to save/restore
596 the unget state (explained later).
600 if (is_block_comment(tk.txt)) {
603 reset_token(state, &tk);
606 save_unget_state(state);
607 ch = get_char(state);
609 while (!at_eon(state) &&
610 (prev != '/' || ch != '*') &&
611 (prev != '*' || ch != '/')) {
615 save_unget_state(state);
616 ch = get_char(state);
618 close_token(state, &tk);
624 /* embedded. Need to unget twice! */
625 restore_unget_state(state);
630 tk.num = TK_block_comment;
631 if (newlines && !(ignored & (1<<TK_newline))) {
632 /* next char must be newline */
633 ch = get_char(state);
638 if (tk.num == TK_error ||
639 !(ignored & (1 << TK_block_comment)))
644 ### Indents, Newlines, and White Space.
646 Normally white space is ignored. However newlines can be important as
647 can indents, which are either after a newline or at the start of a
648 node (detected by `at_son()`);
650 ###### exported functions
651 static inline int is_newline(wchar_t ch)
653 return ch == '\n' || ch == '\f' || ch == '\v';
657 if (ch <= ' ' && !is_newline(ch)
661 If a line starts with more white-space than the previous non-blank
662 line - or if the first non-blank line in the document starts with any
663 white-space - then an "IN" is reported at the start of the line.
665 Before the next non-blank line which starts with less white space, or
666 at the latest at the end of the document, a matching "OUT" token
667 is reported. There will always be an exact match between "IN" and
670 It is possible for "OUT" to be followed (almost) immediately by an
671 "IN". This happens if, for example, the indent of three consecutive
672 lines are 0, 8, 4 spaces. Before the second line we report an
673 "IN". Before the third line we must report an "OUT", as 4 is less
674 than 8, then also an Ident as 4 is greater than 0.
680 For the purpose of measuring the length of white space, a tab adds at
681 least one space, and rounds up to a multiple of 8.
683 ###### exported functions
684 static inline int indent_tab(int indent)
689 We need to track the current levels of indent. This requires some
690 sort of stack as indent levels are pushed on and popped off. In
691 practice this stack is unlikely to often exceed 5 so we will used a
692 fixed stack of 20 indent levels. More than this will be silently
697 int indent_sizes[20];
701 Newlines can optionally be reported. Newlines within a block comment
702 or a multi-line string are not reported separately, but each of these
703 must be followed immediately by a newline so these constructs cannot
704 hide the fact that a newline was present.
706 When indents are being reported, the Newline which would normally be
707 reported immediately before the "IN" is delayed until after the
708 matching "OUT". This makes an indented section act like a
709 continuation of the previous line to some extent.
711 A blank line would normally be reported simply as two consecutive Newline
712 tokens. However if the subsequent line is indented (and indents are being
713 reported) then the right thing to do is less obvious as Newlines should be
714 delayed - but how many Newlines?
716 The approach we will take is to report the extra Newlines immediately after
717 the IN token, so the blank line is treated as though it were an indented
723 If we find a newline or white space at the start of a block, we keep
724 collecting spaces, tabs, and newlines until we find some real text.
725 Then depending on the indent we generate some number of tokens. These
726 will be a sequence of "Newline OUT" pairs representing a decrease
727 in indent, then either a Newline or an IN depending on whether the
728 next line is indented, then zero or more Newlines representing all the
729 blank lines that have been skipped.
731 When a Newline leads to the next block of code there is a question of
732 whether the various Newline and OUT/IN tokens should appear to
733 pbelong to the earlier or later block. This is addressed by processing
734 the tokens in two stages based on the relative indent levels of the
735 two blocks (each block has a base indent to which the actual indents
738 Any "Newline OUT" pairs needed to reduce the current indent to the
739 maximum of the base indents of the old and new blocks are generated
740 against the old block. Then if the next block does not have an
741 increased indent, one more "Newline" is generated.
743 If further "Newline OUT" pairs are needed to get to the indent
744 level of the 'next' block, they are generated against that block,
745 though the first Newline is suppressed (it having already been
748 Finally the Newline or IN for the first line of the new block is
749 generated, unless the Newline needs to be suppressed because it
750 appeared at the end of the previous block.
752 This means that a block may start with an OUT or an IN, but
753 will only start with a Newline if it actually starts with a blank
756 We will need to represent in the `token_state` where in this sequence
757 of delayed tokens we are. As `state.col` records the target indent we
758 don't need to record how many OUTs or INs are needed. We do
759 need to record the number of blank lines, and which of Newline and
760 OUT is needed next in the initial sequence of pairs.
762 For this we store one more than the number of blank lines as
763 `delayed_lines` and a flag for `out_next`.
770 Generating these tokens involve two separate pieces of code.
772 Firstly we need to recognise white space and count the indents and
773 newlines. These are recorded in the above state fields.
775 Separately we need, on each call to `token_next`, we need to check if
776 there are some delayed tokens and if so we need to advance the state
777 information and return one token.
780 if (is_newline(ch) || (at_son(state) && ch <= ' ')) {
782 int was_son = at_son(state);
783 if (ignored & (1<<TK_in)) {
786 if (ignored & (1<<TK_newline))
789 close_token(state, &tk);
792 // Indents are needed, so check all white space.
793 while (ch <= ' ' && !at_eon(state)) {
796 ch = get_char(state);
800 if (state->node->next &&
801 state->node->next->indent > state->node->indent)
802 state->col = state->node->next->indent;
804 state->col = state->node->indent;
807 state->delayed_lines = newlines;
808 state->out_next = was_son;
809 state->check_indent = 1;
814 ###### delayed tokens
816 if (state->check_indent || state->delayed_lines) {
817 if (state->col < state->indent_sizes[state->indent_level]) {
818 if (!state->out_next &&
819 !(ignored & (1<<TK_newline))) {
824 state->indent_level -= 1;
829 if (state->col > state->indent_sizes[state->indent_level] &&
830 state->indent_level < sizeof(state->indent_sizes)-1) {
831 state->indent_level += 1;
832 state->indent_sizes[state->indent_level] = state->col;
833 state->delayed_lines -= 1;
837 state->check_indent = 0;
838 if (state->delayed_lines && !(ignored & (1<<TK_newline))) {
840 state->delayed_lines -= 1;
843 state->delayed_lines = 0;
849 After the last newline in the file has been processed, a special
850 end-of-file token will be returned. any further attempts to get more
851 tokens will continue to return the same end-of-file token.
861 state->check_indent = 1;
868 ### Unknown Marks, or errors.
870 We have now handled all the possible known mark-like tokens.
871 If the token we have is not empty and `TK_mark` is allowed,
872 we have an unknown mark, otherwise this must be an error.
875 /* one unknown character */
876 close_token(state, &tk);
880 ## Tools For The Task
882 You may have noticed that are few gaps we left in the above -
883 functions used without first defining them. Doing so above would have
886 ### Character by character
888 As we walk through the various `code_node`s we need to process whole
889 Unicode codepoints, and keep track of which line and column we are on.
890 We will assume for now that any printing character uses one column,
891 though that is not true in general.
893 As the text in a `code_node` may include an indent that identifies it as
894 being code, we need to be careful to strip that. The `code_node` has
895 a flag that tells us whether or not we need to strip.
901 struct code_node *node;
906 ###### internal functions
908 static int do_strip(struct token_state *state)
911 if (state->node->needs_strip) {
913 while (n && state->node->code.txt[state->offset] == ' ') {
918 while (n == 4 && state->node->code.txt[state->offset] == '\t') {
919 indent = indent_tab(indent);
927 static wint_t get_char(struct token_state *state)
933 if (state->node == NULL)
935 if (state->node->code.len <= state->offset) {
937 state->node = state->node->next;
938 while (state->node && state->node->code.txt == NULL);
940 if (state->node == NULL)
942 state->line = state->node->line_no;
943 state->col = do_strip(state);
948 memset(&mbstate, 0, sizeof(mbstate));
950 n = mbrtowc(&next, state->node->code.txt + state->offset,
951 state->node->code.len - state->offset,
953 if (n == -2 || n == 0) {
954 /* Not enough bytes - not really possible */
956 state->offset = state->node->code.len;
957 } else if (n == -1) {
960 next = 0x7f; // an illegal character
966 } else if (is_newline(next)) {
968 state->col = do_strip(state);
969 } else if (next == '\t') {
970 state->col = indent_tab(state->col);
975 We will sometimes want to "unget" the last character as it needs to be
976 considered again as part of the next token. So we need to store a
977 'previous' version of all metadata.
984 ###### before get_char
985 state->prev_offset = state->offset;
986 state->prev_line = state->line;
987 state->prev_col = state->col;
989 ###### internal functions
991 static void unget_char(struct token_state *state)
994 state->offset = state->prev_offset;
995 state->line = state->prev_line;
996 state->col = state->prev_col;
1000 We occasionally need a double-unget, particularly for numbers and
1001 block comments. We don't impose this cost on all scanning, but
1002 require those code sections that need it to call `save_unget_state`
1003 before each `get_char`, and then `restore_unget_state` when a
1004 double-unget is needed.
1011 ###### internal functions
1012 static void save_unget_state(struct token_state *state)
1014 state->prev_offset2 = state->prev_offset;
1015 state->prev_line2 = state->prev_line;
1016 state->prev_col2 = state->prev_col;
1019 static void restore_unget_state(struct token_state *state)
1021 state->prev_offset = state->prev_offset2;
1022 state->prev_line = state->prev_line2;
1023 state->prev_col = state->prev_col2;
1026 At the start of a token we don't want to be at the end of a code block
1027 if we can help it. To avoid this possibility, we 'get' and 'unget' a
1028 single character. This will move into the next non-empty code block
1029 and leave the current pointer at the start of it.
1031 This has to happen _after_ dealing with delayed tokens as some of them
1032 must appear in the previous node. When we do this, we need to reset
1033 the data in the token.
1035 ###### delayed tokens
1036 if (at_eon(state)) {
1039 tk.node = state->node;
1041 tk.txt.txt = state->node->code.txt + state->offset;
1042 tk.line = state->line;
1043 tk.col = state->col;
1049 The current token is initialized to line up with the first character
1050 that we 'get' for each token. When we have, or might have, a full
1051 token we can call `close_token` to set the `len` of the token
1052 appropriately. This can safely be called multiple times.
1054 Finally we occasionally (for single-line strings and block comments)
1055 need to reset to the beginning of the current token as we might have
1056 parsed too much already. For that there is `reset_token`.
1059 tk.node = state->node;
1061 tk.txt.txt = state->node->code.txt + state->offset;
1062 tk.line = state->line;
1063 tk.col = state->col;
1066 ###### internal functions
1068 static void close_token(struct token_state *state,
1071 if (state->node != tk->node)
1072 tk->txt.len = tk->node->code.len - (tk->txt.txt - tk->node->code.txt);
1074 tk->txt.len = (state->node->code.txt + state->offset)
1078 static void reset_token(struct token_state *state, struct token *tok)
1080 state->prev_line = tok->line;
1081 state->prev_col = tok->col;
1082 state->prev_offset = tok->txt.txt - state->node->code.txt;
1088 Tokens make not cross into the next `code_node`, and some tokens can
1089 include the newline at the and of a `code_node`, we must be able to
1090 easily check if we have reached the end. Equally we need to know if
1091 we are at the start of a node, as white space is treated a little
1094 ###### internal functions
1096 static int at_son(struct token_state *state)
1098 return state->offset == 0;
1101 static int at_eon(struct token_state *state)
1103 // at end-of-node ??
1104 return state->node == NULL ||
1105 state->offset >= state->node->code.len;
1108 ### Find a known word
1110 As the known-word list is sorted we can use a simple binary search.
1111 Following the pattern established in "mdcode", we will use a `struct
1112 text` with start and length to represent the code fragment we are
1115 ###### internal functions
1116 static int find_known(struct token_config *conf, struct text txt)
1119 int hi = conf->known_count;
1121 while (lo + 1 < hi) {
1122 int mid = (lo + hi) / 2;
1123 int cmp = strncmp(conf->words_marks[mid],
1125 if (cmp == 0 && conf->words_marks[mid][txt.len])
1132 if (strncmp(conf->words_marks[lo],
1133 txt.txt, txt.len) == 0
1134 && conf->words_marks[lo][txt.len] == 0)
1140 ### Bringing it all together
1142 Now we have all the bits there is just one section missing: combining
1143 all the token parsing code into one block.
1145 The handling of delayed tokens (Newlines, INs, OUTs) must come
1146 first before we try getting another character.
1148 Then we parse all the test, making sure that we check for known marks
1149 before strings and comments, but unknown marks after strings and comments.
1151 This block of code will either return a token, or will choose to
1152 ignore one, in which case it will `continue` around to the top of the
1158 ch = get_char(state);
1167 As well as getting tokens, we need to be able to create the
1168 `token_state` to start with, and discard it later.
1173 ###### main functions
1174 struct token_state *token_open(struct code_node *code, struct
1177 struct token_state *state = malloc(sizeof(*state));
1178 memset(state, 0, sizeof(*state));
1180 state->line = code->line_no;
1181 state->col = do_strip(state);
1185 void token_close(struct token_state *state)
1190 ###### exported functions
1191 struct token_state *token_open(struct code_node *code, struct
1192 token_config *conf);
1193 void token_close(struct token_state *state);
1197 Getting tokens is the main thing but it is also useful to be able to
1198 print out token information, particularly for tracing and testing.
1200 Known tokens are printed verbatim. Other tokens are printed as
1201 `type(content)` where content is truncated to a given number of characters.
1203 The function for printing a truncated string (`text_dump`) is also exported
1204 so that it can be used to tracing processed strings too.
1209 ###### exported functions
1210 void token_trace(FILE *f, struct token tok, int max);
1211 void text_dump(FILE *f, struct text t, int max);
1213 ###### main functions
1215 void text_dump(FILE *f, struct text txt, int max)
1222 for (i = 0; i < max; i++) {
1223 char c = txt.txt[i];
1224 if (c < ' ' || c > '~')
1225 fprintf(f, "\\x%02x", c & 0xff);
1229 fprintf(f, "%c", c);
1235 void token_trace(FILE *f, struct token tok, int max)
1237 static char *types[] = {
1238 [TK_ident] = "ident",
1240 [TK_number] = "number",
1241 [TK_string] = "string",
1242 [TK_multi_string] = "mstring",
1243 [TK_line_comment] = "lcomment",
1244 [TK_block_comment] = "bcomment",
1247 [TK_newline] = "newline",
1249 [TK_error] = "ERROR",
1253 default: /* known word or mark */
1254 fprintf(f, "%.*s", tok.txt.len, tok.txt.txt);
1260 /* No token text included */
1261 fprintf(f, "%s()", types[tok.num]);
1267 case TK_multi_string:
1268 case TK_line_comment:
1269 case TK_block_comment:
1271 fprintf(f, "%s(", types[tok.num]);
1272 text_dump(f, tok.txt, max);
1278 ### And there we have it
1280 We now have all the library functions defined for reading and printing
1281 tokens. Now we just need C files to store them, and a mk file to make them.
1283 ###### File: scanner.h
1285 ## exported functions
1287 ###### File: libscanner.c
1289 #include "scanner.h"
1291 ## internal functions
1294 ###### File: scanner.mk
1298 scanner.mk scanner.h libscanner.c : scanner.mdc
1301 libscanner.o : libscanner.c
1302 $(CC) $(CFLAGS) -c libscanner.c
1304 ## Processing numbers
1306 Converting a `TK_number` token to a numerical value is a slightly
1307 higher level task than lexical analysis, and slightly lower than
1308 grammar parsing, so put it here - as an index if you like.
1310 Importantly it will be used by the same testing rig that is used for
1311 testing the token scanner.
1313 The numeric value that we will convert all numbers into is the `mpq_t`
1314 from the GNU high precision number library "libgmp".
1316 ###### number includes
1320 Firstly we need to be able to parse a string of digits in a given base
1321 and possibly with a decimal marker. We store this in an `mpz_t`
1322 integer and report the number of digits after the decimal mark.
1324 On error we return zero and ensure that the 'mpz_t' has been freed, or
1325 had never been initialised.
1327 ###### number functions
1329 static int parse_digits(mpz_t num, struct text tok, int base,
1332 /* Accept digits up to 'base', ignore '_' and
1333 * ' ' if they appear between two legal digits,
1334 * and if `placesp` is not NULL, allow a single
1335 * '.' or ',' and report the number of digits
1337 * Return number of characters processed (p),
1338 * or 0 if something illegal was found.
1341 int decimal = -1; // digits after marker
1342 enum {Digit, Space, Other} prev = Other;
1345 for (p = 0; p < tok.len; p++) {
1347 char c = tok.txt[p];
1349 if (c == '_' || c == ' ') {
1355 if (c == '.' || c == ',') {
1358 if (!placesp || decimal >= 0)
1366 else if (isupper(c))
1368 else if (islower(c))
1379 mpz_mul_ui(num, num, base);
1383 mpz_add_ui(num, num, dig);
1402 ###### number includes
1405 To parse a full number we need to consider the optional base, the
1406 mantissa, and the optional exponent. We will treat these one at a
1409 The base is indicated by a letter after a leading zero, which must be
1410 followed by a base letter or a period. The base also determines the
1411 character which will mark an exponent.
1419 if (tok.txt[0] == '0' && tok.len > 1) {
1421 switch(tok.txt[1]) {
1452 // another digit is not permitted
1456 // must be decimal marker or trailing
1457 // letter, which are OK;
1464 After the base is the mantissa, which may contain a decimal mark, so
1465 we need to record the number of places. We won't impose the number of
1466 places until we have the exponent as well.
1473 ###### parse mantissa
1475 d = parse_digits(mant, tok, base, &places);
1481 mpq_set_z(num, mant);
1484 After the mantissa number may come an exponent which may be positive
1485 or negative. We assume at this point that we have seen the exponent
1493 ###### parse exponent
1495 if (tok.txt[0] == '+') {
1498 } else if (tok.txt[0] == '-') {
1504 d = parse_digits(exp, tok, 10, NULL);
1509 if (!mpz_fits_slong_p(exp)) {
1514 lexp = mpz_get_si(exp) * esign;
1520 Now that we have the mantissa and the exponent we can multiply them
1521 together, also allowing for the number of digits after the decimal
1524 For base 10, we simply subtract the decimal places from the exponent.
1525 For the other bases, as the exponent is alway based on 2, even for
1526 octal and hex, we need a bit more detail.
1527 We then recover the sign from the exponent, as division is quite
1528 different from multiplication.
1530 ###### calc exponent
1549 Imposing the exponent on the number is also very different for base 10
1550 than for the others. For the binary shift `gmp` provides a simple
1551 function. For base 10 we use something like Russian Peasant
1554 ###### calc exponent
1558 mpq_set_ui(tens, 10, 1);
1562 mpq_mul(num, num, tens);
1564 mpq_div(num, num, tens);
1569 mpq_mul(tens, tens, tens);
1574 mpq_mul_2exp(num, num, lexp);
1576 mpq_div_2exp(num, num, lexp);
1579 Now we are ready to parse a number: the base, mantissa, and exponent.
1580 If all goes well we check for the possible trailing letters and
1581 return. Return value is 1 for success and 0 for failure.
1584 ###### number functions
1585 int number_parse(mpq_t num, char tail[3], struct text tok)
1592 if (tok.len > 1 && (tok.txt[0] == expc ||
1593 tok.txt[0] == toupper(expc))) {
1600 for (i = 0; i < 2; i++) {
1603 if (!isalpha(tok.txt[i]))
1605 tail[i] = tok.txt[i];
1615 Number parsing goes in `libnumber.c`
1617 ###### File: libnumber.c
1625 ###### File: number.h
1626 int number_parse(mpq_t num, char tail[3], struct text tok);
1628 ###### File: scanner.mk
1630 libnumber.o : libnumber.c
1631 $(CC) $(CFLAGS) -c libnumber.c
1633 ## Processing strings
1635 Both `TK_string` and `TK_multi_string` require post-processing which
1636 can be one of two types: literal or with escapes processed.
1637 Even literal processing is non-trivial as the file may contain indents
1638 which need to be stripped.
1640 Errors can only occur when processing escapes. Any unrecognised
1641 character following the escape character will cause an error.
1643 Processing escapes and striping indents can only make the string
1644 shorter, not longer, so we allocate a buffer which is the same size as
1645 the string and process into that.
1647 To request escape processing, we pass the character we want to use for
1648 quoting, usually '`\`'. To avoid escape processing we pass a zero.
1651 int string_parse(struct token *tok, char escape,
1652 struct text *str, char tail[3])
1655 struct text t = tok->txt;
1659 if (tok->num == TK_string) {
1664 str->txt = malloc(t.len);
1677 The tail of the string can be 0, 1, or 2 letters
1680 if (i >= 0 && isalpha(t.txt[i-1]))
1682 if (i >= 0 && isalpha(t.txt[i-1]))
1684 strncpy(tail, t.txt+i, t.len-i);
1693 Stripping the quote of a single-line string is trivial.
1694 The only part that is at all interesting is that quote character must
1698 if (t.txt[t.len-1] != quote)
1708 For a multi-line string we have a little more work to do. We need to
1709 remove 3 quotes, not 1, and need to count the indent of the close
1710 quote as it will need to be stripped from all lines.
1714 t.txt[1] != quote || t.txt[2] != quote ||
1715 !is_newline(t.txt[3]))
1720 if (i <= 0 || t.txt[i-1] != quote)
1723 if (i <= 0 || t.txt[i-1] != quote)
1726 if (i <= 0 || t.txt[i-1] != quote)
1730 while (i > 0 && !is_newline(t.txt[i-1]))
1734 if (t.txt[i] == ' ')
1736 if (t.txt[i] == '\t')
1737 indent = indent_tab(indent);
1746 Now we just take one byte at a time. trans-ASCII unicode won't look
1747 like anything we are interested in so it will just be copied byte by
1752 for (i = 0; i < t.len; i++) {
1766 } else if (i+1 >= t.len) {
1767 // escape and end of string
1775 str->len = cp - str->txt;
1783 Every time we find a start of line, we strip spaces and tabs until the
1784 required indent is found.
1787 while (i < t.len && skipped < indent) {
1792 skipped = indent_tab(skipped);
1801 *cp++ = '\n'; break;
1803 *cp++ = '\r'; break;
1805 *cp++ = '\t'; break;
1807 *cp++ = '\b'; break;
1809 *cp++ = quote; break;
1811 *cp++ = '\f'; break;
1813 *cp++ = '\v'; break;
1815 *cp++ = '\a'; break;
1820 // 3 digit octal number
1823 if (t.txt[i+1] < '0' || t.txt[i+1] > '7' ||
1824 t.txt[i+2] < '0' || t.txt[i+1] > '7')
1826 n = (t.txt[i ]-'0') * 64 +
1827 (t.txt[i+1]-'0') * 8 +
1828 (t.txt[i+2]-'0') * 1;
1834 n = take_hex(2, t.txt+i+1, t.len-i-1);
1842 // 4 or 8 hex digits for unicode
1843 n = take_hex(c == 'u'?4:8, t.txt+i+1, t.len-i-1);
1846 memset(&pstate, 0, sizeof(pstate));
1847 n = wcrtomb(cp, n, &pstate);
1851 i += c == 'u' ? 4 : 8;
1856 else if (is_newline(c))
1866 For `\x` `\u` and `\U` we need to collect a specific number of
1869 ###### string functions
1871 static long take_hex(int digits, char *cp, int l)
1883 else if (isupper(c))
1894 #### File: libstring.c
1896 String parsing goes in `libstring.c`
1905 #include "scanner.h"
1909 ###### File: string.h
1910 int string_parse(struct token *tok, char escape,
1911 struct text *str, char tail[3]);
1913 ###### File: scanner.mk
1915 libstring.o : libstring.c
1916 $(CC) $(CFLAGS) -c libstring.c
1921 As "untested code is buggy code" we need a program to easily test
1922 the scanner library. This will simply parse a given file and report
1923 the tokens one per line.
1925 ###### File: scanner.c
1931 #include <sys/mman.h>
1937 #include "scanner.h"
1942 static void pr_err(char *msg)
1945 fprintf(stderr, "%s\n", msg);
1948 int main(int argc, char *argv[])
1953 struct token_state *state;
1954 const char *known[] = {
1963 struct token_config conf = {
1966 .words_marks = known,
1967 .number_chars = "., _+-",
1968 .known_count = sizeof(known)/sizeof(known[0]),
1969 .ignored = (0 << TK_line_comment)
1970 |(0 << TK_block_comment),
1972 struct section *table, *s, *prev;
1973 setlocale(LC_ALL,"");
1975 fprintf(stderr, "Usage: scanner file\n");
1978 fd = open(argv[1], O_RDONLY);
1980 fprintf(stderr, "scanner: cannot open %s: %s\n",
1981 argv[1], strerror(errno));
1984 len = lseek(fd, 0, 2);
1985 file = mmap(NULL, len, PROT_READ, MAP_SHARED, fd, 0);
1986 table = code_extract(file, file+len, pr_err);
1989 (code_free(s->code), prev = s, s = s->next, free(prev))) {
1990 printf("Tokenizing: %.*s\n", s->section.len,
1992 state = token_open(s->code, &conf);
1994 struct token tk = token_next(state);
1995 printf("%d:%d ", tk.line, tk.col);
1996 token_trace(stdout, tk, 20);
1997 if (tk.num == TK_number) {
2000 if (number_parse(num, tail,tk.txt)) {
2001 printf(" %s ", tail);
2002 mpq_out_str(stdout, 10, num);
2005 printf(" BAD NUMBER");
2007 if (tk.num == TK_string ||
2008 tk.num == TK_multi_string) {
2012 if (tk.txt.txt[0] == '`')
2014 if (string_parse(&tk, esc,
2016 printf(" %s ", tail);
2017 text_dump(stdout, str, 20);
2020 printf(" BAD STRING");
2023 if (tk.num == TK_error)
2025 if (tk.num == TK_eof)
2031 ###### File: scanner.mk
2032 scanner.c : scanner.mdc
2035 scanner : scanner.o scanner.h libscanner.o libmdcode.o mdcode.h
2036 $(CC) $(CFLAGS) -o scanner scanner.o libscanner.o \
2037 libmdcode.o libnumber.o libstring.o -licuuc -lgmp
2038 scanner.o : scanner.c
2039 $(CC) $(CFLAGS) -c scanner.c