1

/*

     2

 * (c) 2003 Mario de Sousa

     3

 *

     4

 * Offered to the public under the terms of the GNU General Public License

     5

 * as published by the Free Software Foundation; either version 2 of the

     6

 * License, or (at your option) any later version.

     7

 *

     8

 * This program is distributed in the hope that it will be useful, but

     9

 * WITHOUT ANY WARRANTY; without even the implied warranty of

    10

 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General

    11

 * Public License for more details.

    12

 *

    13

 * This code is made available on the understanding that it will not be

    14

 * used in safety-critical situations without a full and competent review.

    15

 */

    16

    17

/*

    18

 * An IEC 61131-3 IL and ST compiler.

    19

 *

    20

 * Based on the

    21

 * FINAL DRAFT - IEC 61131-3, 2nd Ed. (2001-12-10)

    22

 *

    23

 */

    24

    25

/*

    26

 * Stage 1

    27

 * =======

    28

 *

    29

 * This file contains the lexical tokens definitions, from which

    30

 * the flex utility will generate a lexical parser function.

    31

 */

    32

    33

    34

    35

    36

/*****************************/

    37

/* Lexical Parser Options... */

    38

/*****************************/

    39

    40

/* The lexical analyser will never work in interactive mode,

    41

 * i.e., it will only process programs saved to files, and never

    42

 * programs being written inter-actively by the user.

    43

 * This option saves the resulting parser from calling the

    44

 * isatty() function, that seems to be generating some compile

    45

 * errors under some (older?) versions of flex.

    46

 */

    47

%option never-interactive

    48

    49

/* Have the lexical analyser use a 'char *yytext' instead of an

    50

 * array of char 'char yytext[??]' to store the lexical token.

    51

 */

    52

%pointer

    53

    54

    55

/* Have the lexical analyser ignore the case of letters.

    56

 * This will occur for all the tokens and keywords, but

    57

 * the resulting text handed up to the syntax parser

    58

 * will not be changed, and keep the original case

    59

 * of the letters in the input file.

    60

 */

    61

%option case-insensitive

    62

    63

/* Have the generated lexical analyser keep track of the

    64

 * line number it is currently analysing.

    65

 * This is used to pass up to the syntax parser

    66

 * the number of the line on which the current

    67

 * token was found. It will enable the syntax parser

    68

 * to generate more informatve error messages...

    69

 */

    70

%option yylineno

    71

    72

/* required for the use of the yy_pop_state() and

    73

 * yy_push_state() functions

    74

 */

    75

%option stack

    76

    77

/* The '%option stack' also requests the inclusion of 

    78

 * the yy_top_state(), however this function is not

    79

 * currently being used. This means that the compiler

    80

 * is complaining about the existance of this function.

    81

 * The following option removes the yy_top_state()

    82

 * function from the resulting c code, so the compiler 

    83

 * no longer complains.

    84

 */

    85

%option noyy_top_state

    86

    87

/**************************************************/

    88

/* External Variable and Function declarations... */

    89

/**************************************************/

    90

    91

    92

%{

    93

/* Define TEST_MAIN to include a main() function.

    94

 * Useful for testing the parser generated by flex.

    95

 */

    96

/*

    97

#define TEST_MAIN

    98

*/

    99

/* If lexical parser is compiled by itself, we need to define the following

   100

 * constant to some string. Under normal circumstances LIBDIRECTORY is set

   101

 * in the syntax parser header file...

   102

 */

   103

#ifdef TEST_MAIN

   104

#define LIBDIRECTORY "just_testing"

   105

#endif

   106

   107

   108

   109

/* Required for strdup() */

   110

#include <string.h>

   111

   112

/* Required only for the declaration of abstract syntax classes

   113

 * (class symbol_c; class token_c; class list_c;)

   114

 * These will not be used in flex, but the token type union defined

   115

 * in iec.hh contains pointers to these classes, so we must include

   116

 * it here.

   117

 */

   118

#include "../absyntax/absyntax.hh"

   119

   120

/* generated by bison.

   121

 * Contains the definition of the token constants, and the

   122

 * token value type YYSTYPE (in our case, a 'const char *')

   123

 */

   124

#include "iec.y.hh"

   125

   126

/* Variable defined by the bison parser,

   127

 * where the value of the tokens will be stored

   128

 */

   129

extern YYSTYPE yylval;

   130

   131

/* The name of the file currently being parsed...

   132

 * This variable is declared and read from the code generated by bison!

   133

 * Note that flex accesses and updates this global variable

   134

 * apropriately whenever it comes across an (*#include <filename> *)

   135

 * directive...

   136

 */

   137

extern const char *current_filename;

   138

   139

/* We will not be using unput() in our flex code... */

   140

#define YY_NO_UNPUT

   141

   142

/* Variable defined by the bison parser.

   143

 * It must be initialised with the location

   144

 * of the token being parsed.

   145

 * This is only needed if we want to keep

   146

 * track of the locations, in order to give

   147

 * more meaningful error messages!

   148

 */

   149

extern YYLTYPE yylloc;

   150

   151

/* Macro that is executed for every action.

   152

 * We use it to pass the location of the token

   153

 * back to the bison parser...

   154

 */

   155

#define YY_USER_ACTION { 					\

   156

	yylloc.first_line = yylloc.last_line = yylineno;	\

   157

	yylloc.first_column = yylloc.last_column = 0;		\

   158

	}

   159

   160

   161

/* Since this lexical parser we defined only works in ASCII based

   162

 * systems, we might as well make sure it is being compiled on

   163

 * one...

   164

 * Lets check a few random characters...

   165

 */

   166

#if (('a' != 0x61) || ('A' != 0x41) || ('z' != 0x7A) || ('Z' != 0x5A) || \

   167

     ('0' != 0x30) || ('9' != 0x39) || ('(' != 0x28) || ('[' != 0x5B))

   168

#error This lexical analyser is not portable to a non ASCII based system.

   169

#endif

   170

   171

   172

/* Function only called from within flex, but defined

   173

 * in iec.y!

   174

 * We declare it here...

   175

 *

   176

 * Search for a symbol in either of the two symbol tables

   177

 * and return the token id of the first symbol found.

   178

 * Searches first in the variables, and only if not found

   179

 * does it continue searching in the library elements

   180

 */

   181

//token_id_t get_identifier_token(const char *identifier_str);

   182

int get_identifier_token(const char *identifier_str);

   183

%}

   184

   185

   186

/*********************************************/

   187

/* Change parse state to body definitions... */

   188

/*********************************************/

   189

   190

%{

   191

static int goto_body_state__ = 0;

   192

   193

void cmd_goto_body_state(void) {goto_body_state__ = 1;}

   194

int  get_goto_body_state(void) {return goto_body_state__;}

   195

void rst_goto_body_state(void) {goto_body_state__ = 0;}

   196

%}

   197

   198

   199

/***************************************************/

   200

/* Forward Declaration of functions defined later. */

   201

/***************************************************/

   202

   203

%{

   204

/* return all the text in the current token back to the input stream. */

   205

void unput_text(unsigned int n);

   206

%}

   207

   208

   209

   210

/****************************/

   211

/* Lexical Parser States... */

   212

/****************************/

   213

   214

/* NOTE: Our psrser can parse st or il code, intermixed

   215

 *       within the same file.

   216

 *       With IL we come across the issue of the EOL (end of line) token.

   217

 *       ST, and the declaration parts of IL do not use this token!

   218

 *       If the lexical analyser were to issue this token during ST

   219

 *       language parsing, or during the declaration of data types,

   220

 *       function headers, etc. in IL, the syntax parser would crash.

   221

 *

   222

 *       We can solve this issue using one of three methods:

   223

 *        (1) Augment all the syntax that does not accept the EOL

   224

 *            token to simply ignore it. This makes the syntax

   225

 *            definition (in iec.y) very cluttered!

   226

 *        (2) Let the lexical parser figure out which language

   227

 *            it is parsing, and decide whether or not to issue

   228

 *            the EOL token. This requires the lexical parser

   229

 *            to have knowledge of the syntax!, making for a poor

   230

 *            overall organisation of the code. It would also make it

   231

 *            very difficult to understand the lexical parser as it

   232

 *            would use several states, and a state machine to transition

   233

 *            between the states. The state transitions would be

   234

 *            intermingled with the lexical parser defintion!

   235

 *        (3) Use a mixture of (1) and (2). The lexical analyser

   236

 *            merely distinguishes between function headers and function

   237

 *            bodies, but no longer makes a distinction between il and

   238

 *            st language bodies. When parsing a body, it will return

   239

 *            the EOL token. In other states '\n' will be ignored as

   240

 *            whitespace.

   241

 *            The ST language syntax has been augmented in the syntax

   242

 *            parser configuration to ignore any EOL tokens that it may

   243

 *            come across!

   244

 *            This option has both drawbacks of option (1) and (2), but

   245

 *            much less intensely.

   246

 *            The syntax that gets cluttered is limited to the ST statements

   247

 *            (which is rather limited, compared to the function headers and

   248

 *            data type declarations, etc...), while the state machine in

   249

 *            the lexical parser becomes very simple. All state transitions

   250

 *            can be handled within the lexical parser by itself, and can be

   251

 *            easily identified. Thus knowledge of the syntax required by

   252

 *            the lexical parser is very limited!

   253

 *

   254

 * Amazingly enough, I (Mario) got to implement option (3)

   255

 * at first, requiring two basic states, decl and body.

   256

 * The lexical parser will enter the body state when

   257

 * it is parsing the body of a function/program/function block. The

   258

 * state transition is done when we find a VAR_END that is not followed

   259

 * by a VAR! This is the syntax knowledge that gets included in the

   260

 * lexical analyser with this option!

   261

 * Unfortunately, getting the st syntax parser to ignore EOL anywhere

   262

 * where they might appear leads to conflicts. This is due to the fact

   263

 * that the syntax parser uses the single look-ahead token to remove

   264

 * possible conflicts. When we insert a possible EOL, the single

   265

 * look ahead token becomes the EOL, which means the potential conflicts

   266

 * could no longer be resolved.

   267

 * Removing these conflicts would make the st syntax parser very convoluted,

   268

 * and adding the extraneous EOL would make it very cluttered.

   269

 * This option was therefore dropped in favour of another!

   270

 *

   271

 * I ended up implementing (2). Unfortunately the lexical analyser can

   272

 * not easily distinguish between il and st code, since function

   273

 * calls in il are very similar to function block calls in st.

   274

 * We therefore use an extra 'body' state. When the lexical parser

   275

 * finds that last END_VAR, it enters the body state. This state

   276

 * must figure out what language is being parsed from the first few

   277

 * tokens, and switch to the correct state (st or il) according to the

   278

 * language. This means that we insert quite a bit of knowledge of the

   279

 * syntax of the languages into the lexical parser. This is ugly, but it

   280

 * works, and at least it is possible to keep all the state changes together

   281

 * to make it easier to remove them later on if need be.

   282

 * The body state returns any matched text back to the buffer with unput(),

   283

 * to be later matched correctly by the apropriate language parser (st or il).

   284

 * The state machine has 6 possible states (INITIAL, config, decl, body, st, il)

   285

 * Possible state changes are:

   286

 *   INITIAL -> decl_state (when a FUNCTION, FUNCTION_BLOCK, or PROGRAM is found,

   287

 *                    and followed by a VAR declaration)

   288

 *   INITIAL -> il_st_state (when a FUNCTION, FUNCTION_BLOCK, or PROGRAM is found,

   289

 *                    and _not_ followed by a VAR declaration)

   290

 *   INITIAL -> config_state (when a CONFIGURATION is found)

   291

 *   decl_state    -> il_st_state (when the last END_VAR is found, i.e. the function body starts)

   292

 *   il_st_state   -> sfc_state (when it figures out it is parsing sfc language)

   293

 *   il_st_state   -> st_state (when it figures out it is parsing st language)

   294

 *   il_st_state   -> il_state (when it figures out it is parsing il language)

   295

 *   decl_state    -> INITIAL (when a END_FUNCTION, END_FUNCTION_BLOCK, or END_PROGRAM is found)

   296

 *   st_state      -> INITIAL (when a END_FUNCTION, END_FUNCTION_BLOCK, or END_PROGRAM is found)

   297

 *   sfc_state     -> INITIAL (when a END_FUNCTION, END_FUNCTION_BLOCK, or END_PROGRAM is found)

   298

 *   il_state      -> INITIAL (when a END_FUNCTION, END_FUNCTION_BLOCK, or END_PROGRAM is found)

   299

 *   config_state  -> INITIAL (when a END_CONFIGURATION is found)

   300

 */

   301

/* we are parsing a configuration. */

   302

%s config_state

   303

   304

/* we are parsing a function, program or function block declaration */

   305

%s decl_state

   306

   307

/* we will be parsing a function body. Whether il/st is remains unknown */

   308

%x il_st_state

   309

   310

/* we are parsing il code -> flex must return the EOL tokens!       */

   311

%s il_state

   312

   313

/* we are parsing st code -> flex must not return the EOL tokens!   */

   314

%s st_state

   315

   316

/* we are parsing sfc code -> flex must not return the EOL tokens!   */

   317

%s sfc_state

   318

   319

   320

   321

/*******************/

   322

/* File #include's */

   323

/*******************/

   324

   325

/* We extend the IEC 61131-3 standard syntax to allow inclusion

   326

 * of other files, using the IEC 61131-3 pragma directive...

   327

 * The accepted syntax is:

   328

 *  {#include "<filename>"}

   329

 */

   330

   331

/* the "include" states are used for picking up the name of an include file */

   332

%x include_beg

   333

%x include_filename

   334

%x include_end

   335

   336

   337

file_include_pragma_filename	[^\"]*

   338

file_include_pragma_beg		"{#include"{st_whitespace_only}\"

   339

file_include_pragma_end		\"{st_whitespace_only}"}"

   340

file_include_pragma			{file_include_pragma_beg}{file_include_pragma_filename}{file_include_pragma_end}

   341

   342

   343

%{

   344

#define MAX_INCLUDE_DEPTH 16

   345

   346

typedef struct {

   347

	  YY_BUFFER_STATE buffer_state;

   348

	  int lineno;

   349

	  const char *filename;

   350

	} include_stack_t;

   351

   352

include_stack_t include_stack[MAX_INCLUDE_DEPTH];

   353

int include_stack_ptr = 0;

   354

   355

const char *INCLUDE_DIRECTORIES[] = {

   356

	"",

   357

	"lib/",

   358

	"/lib/",

   359

	"/usr/lib/",

   360

	"/usr/lib/iec/",

   361

	LIBDIRECTORY "/",

   362

	NULL /* must end with NULL!! */

   363

	};

   364

   365

   366

/*

   367

 * Join two strings together. Allocate space with malloc(3).

   368

 */

   369

static char *strdup2(const char *a, const char *b) {

   370

  char *res = (char *)malloc(strlen(a) + strlen(b) + 1);

   371

   372

  if (!res)

   373

    return NULL;

   374

  return strcat(strcpy(res, a), b);  /* safe, actually */

   375

}

   376

%}

   377

   378

   379

   380

/*****************************/

   381

/* Prelimenary constructs... */

   382

/*****************************/

   383

   384

   385

/* A pragma... */

   386

   387

pragma "{"[^}]*"}"

   388

   389

/* NOTE: this seemingly unnecessary complex definition is required

   390

 *       to be able to eat up comments such as:

   391

 *          '(* Testing... ! ***** ******)'

   392

 *       without using the trailing context command in flex (/{context})

   393

 *       since {comment} itself will later be used with

   394

 *       trailing context ({comment}/{context})

   395

 */

   396

not_asterisk				[^*]

   397

not_close_parenthesis_nor_asterisk	[^*)]

   398

asterisk				"*"

   399

comment_text		{not_asterisk}|(({asterisk}+){not_close_parenthesis_nor_asterisk})

   400

   401

comment		"(*"({comment_text}*)({asterisk}+)")"

   402

   403

   404

/*

   405

3.1 Whitespace

   406

 (NOTE: Whitespace IS clearly defined, to include newline!!! See section 2.1.4!!!)

   407

 No definition of whitespace is given, in other words, the characters that may be used to seperate language tokens are not pecisely defined. One may nevertheless make an inteligent guess of using the space (' '), and other characters also commonly considered whitespace in other programming languages (horizontal tab, vertical tab, form feed, etc.).

   408

 The main question is whether the newline character should be considered whitespace. IL language statements use an EOL token (End Of Line) to distinguish between some language constructs. The EOL token itself is openly defined as "normally consist[ing] of the 'paragraph separator' ", leaving the final choice open to each implemention. If we choose the newline character to represent the EOL token, it may then not be considered whitespace.

   409

 On the other hand, some examples that come in a non-normative annex of the specification allow function declarations to span multiple3.1 Whitespace

   410

 (NOTE: Whitespace IS clearly defined, to include newline!!! See section 2.1.4!!!)

   411

 No definition of whitespace is given, in other words, the characters that may be used to seperate language tokens are not pecisely defined. One may nevertheless make an inteligent guess of using the space (' '), and other characters also commonly considered whitespace in other programming languages (horizontal tab, vertical tab, form feed, etc.).

   412

 The main question is whether the newline character should be considered whitespace. IL language statements use an EOL token (End Of Line) to distinguish between some language constructs. The EOL token itself is openly defined as "normally consist[ing] of the 'paragraph separator' ", leaving the final choice open to each implemention. If we choose the newline character to represent the EOL token, it may then not be considered whitespace.

   413

 On the other hand, some examples that come in a non-normative annex of the specification allow function declarations to span multiple lines, which means that the newline character is being considered as whitespace.

   414

 Our implementation works around this issue by including the new line character in the whitespace while parsing function declarations and the ST language, and parsing it as the EOL token only while parsing IL language statements. This requires the use of a state machine in the lexical parser that needs at least some knowledge of the syntax itself.

   415

*/

   416

/* NOTE: Our definition of whitespace will only work in ASCII!

   417

 *

   418

 *       Since the IL language needs to know the location of newline

   419

 *       (token EOL -> '\n' ), we need one definition of whitespace

   420

 *       for each language...

   421

 */

   422

/*

   423

 * NOTE: we cannot use

   424

 *         st_whitespace	[:space:]*

   425

 *       since we use {st_whitespace} as trailing context. In our case

   426

 *       this would not constitute "dangerous trailing context", but the

   427

 *       lexical generator (i.e. flex) does not know this (since it does

   428

 *       not know which characters belong to the set [:space:]), and will

   429

 *       generate a "dangerous trailing context" warning!

   430

 *       We use this alternative just to stop the flex utility from

   431

 *       generating the invalid (in this case) warning...

   432

 */

   433

   434

st_whitespace_only	[ \f\n\r\t\v]*

   435

il_whitespace_only	[ \f\r\t\v]*

   436

   437

st_whitespace_text	{st_whitespace_only}|{comment}|{pragma}

   438

il_whitespace_text	{il_whitespace_only}|{comment}|{pragma}

   439

   440

st_whitespace	{st_whitespace_text}*

   441

il_whitespace	{il_whitespace_text}*

   442

   443

st_whitespace_text_no_pragma	{st_whitespace_only}|{comment}

   444

il_whitespace_text_no_pragma	{il_whitespace_only}|{comment}

   445

   446

st_whitespace_no_pragma	{st_whitespace_text_no_pragma}*

   447

il_whitespace_no_pragma	{il_whitespace_text_no_pragma}*

   448

   449

qualified_identifier	{identifier}(\.{identifier})?

   450

   451

   452

   453

/*****************************************/

   454

/* B.1.1 Letters, digits and identifiers */

   455

/*****************************************/

   456

/* NOTE: The following definitions only work if the host computer

   457

 *       is using the ASCII maping. For e.g., with EBCDIC [A-Z]

   458

 *       contains non-alphabetic characters!

   459

 *       The correct way of doing it would be to use

   460

 *       the [:upper:] etc... definitions.

   461

 *

   462

 *       Unfortunately, further on we need all printable

   463

 *       characters (i.e. [:print:]), but excluding '$'.

   464

 *       Flex does not allow sets to be composed by excluding

   465

 *       elements. Sets may only be constructed by adding new

   466

 *       elements, which means that we have to revert to

   467

 *       [\x20\x21\x23\x25\x26\x28-x7E] for the definition

   468

 *       of the printable characters with the required exceptions.

   469

 *       The above also implies the use of ASCII, but now we have

   470

 *       no way to work around it|

   471

 *

   472

 *       The conclusion is that our parser is limited to ASCII

   473

 *       based host computers!!

   474

 */

   475

letter		[A-Za-z]

   476

digit		[0-9]

   477

octal_digit	[0-7]

   478

hex_digit	{digit}|[A-F]

   479

identifier	({letter}|(_({letter}|{digit})))((_?({letter}|{digit}))*)

   480

   481

   482

/*******************/

   483

/* B.1.2 Constants */

   484

/*******************/

   485

   486

/******************************/

   487

/* B.1.2.1   Numeric literals */

   488

/******************************/

   489

integer         {digit}((_?{digit})*)

   490

binary_integer  2#{bit}((_?{bit})*)

   491

bit		[0-1]

   492

octal_integer   8#{octal_digit}((_?{octal_digit})*)

   493

hex_integer     16#{hex_digit}((_?{hex_digit})*)

   494

exponent        [Ee]([+-]?){integer}

   495

/* The correct definition for real would be:

   496

 * real		{integer}\.{integer}({exponent}?)

   497

 *

   498

 * Unfortunately, the spec also defines fixed_point (B 1.2.3.1) as:

   499

 * fixed_point		{integer}\.{integer}

   500

 *

   501

 * This means that {integer}\.{integer} could be interpreted

   502

 * as either a fixed_point or a real.

   503

 * I have opted to interpret {integer}\.{integer} as a fixed_point.

   504

 * In order to do this, the definition of real has been changed to:

   505

 * real		{integer}\.{integer}{exponent}

   506

 *

   507

 * This means that the syntax parser now needs to define a real to be

   508

 * either a real_token or a fixed_point_token!

   509

 */

   510

real		{integer}\.{integer}{exponent}

   511

   512

   513

/*******************************/

   514

/* B.1.2.2   Character Strings */

   515

/*******************************/

   516

/*

   517

common_character_representation :=

   518

<any printable character except '$', '"' or "'">

   519

|'$$'

   520

|'$L'|'$N'|'$P'|'$R'|'$T'

   521

|'$l'|'$n'|'$p'|'$r'|'$t'

   522

   523

NOTE: 	$ = 0x24

   524

	" = 0x22

   525

	' = 0x27

   526