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 DEFAULT_LIBDIR "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

   121

/* generated by bison.

   122

 * Contains the definition of the token constants, and the

   123

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

   124

 */

   125

#include "iec.y.hh"

   126

#include "stage1_2_priv.hh"

   127

   128

   129

/* Variable defined by the bison parser,

   130

 * where the value of the tokens will be stored

   131

 */

   132

extern YYSTYPE yylval;

   133

   134

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

   135

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

   136

 * Note that flex accesses and updates this global variable

   137

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

   138

 * directive...

   139

 */

   140

/*

   141

 NOTE: already defined in iec.y.hh 

   142

extern const char *current_filename;

   143

*/

   144

   145

   146

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

   147

#define YY_NO_UNPUT

   148

   149

/* Variable defined by the bison parser.

   150

 * It must be initialised with the location

   151

 * of the token being parsed.

   152

 * This is only needed if we want to keep

   153

 * track of the locations, in order to give

   154

 * more meaningful error messages!

   155

 */

   156

extern YYLTYPE yylloc;

   157

   158

/* Macro that is executed for every action.

   159

 * We use it to pass the location of the token

   160

 * back to the bison parser...

   161

 */

   162

#define YY_USER_ACTION { 					\

   163

	yylloc.first_line = yylloc.last_line = yylineno;	\

   164

	yylloc.first_column = yylloc.last_column = 0;		\

   165

	}

   166

   167

   168

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

   169

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

   170

 * one...

   171

 * Lets check a few random characters...

   172

 */

   173

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

   174

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

   175

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

   176

#endif

   177

   178

   179

/* Function only called from within flex, but defined

   180

 * in iec.y!

   181

 * We declare it here...

   182

 *

   183

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

   184

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

   185

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

   186

 * does it continue searching in the library elements

   187

 */

   188

//token_id_t get_identifier_token(const char *identifier_str);

   189

int get_identifier_token(const char *identifier_str);

   190

%}

   191

   192

   193

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

   194

/* Forward Declaration of functions defined later. */

   195

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

   196

   197

%{

   198

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

   199

void unput_text(unsigned int n);

   200

%}

   201

   202

   203

   204

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

   205

/* Lexical Parser States... */

   206

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

   207

   208

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

   209

 *       within the same file.

   210

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

   211

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

   212

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

   213

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

   214

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

   215

 *

   216

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

   217

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

   218

 *            token to simply ignore it. This makes the syntax

   219

 *            definition (in iec.y) very cluttered!

   220

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

   221

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

   222

 *            the EOL token. This requires the lexical parser

   223

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

   224

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

   225

 *            very difficult to understand the lexical parser as it

   226

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

   227

 *            between the states. The state transitions would be

   228

 *            intermingled with the lexical parser defintion!

   229

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

   230

 *            merely distinguishes between function headers and function

   231

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

   232

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

   233

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

   234

 *            whitespace.

   235

 *            The ST language syntax has been augmented in the syntax

   236

 *            parser configuration to ignore any EOL tokens that it may

   237

 *            come across!

   238

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

   239

 *            much less intensely.

   240

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

   241

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

   242

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

   243

 *            the lexical parser becomes very simple. All state transitions

   244

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

   245

 *            easily identified. Thus knowledge of the syntax required by

   246

 *            the lexical parser is very limited!

   247

 *

   248

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

   249

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

   250

 * The lexical parser will enter the body state when

   251

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

   252

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

   253

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

   254

 * lexical analyser with this option!

   255

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

   256

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

   257

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

   258

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

   259

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

   260

 * could no longer be resolved.

   261

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

   262

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

   263

 * This option was therefore dropped in favour of another!

   264

 *

   265

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

   266

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

   267

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

   268

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

   269

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

   270

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

   271

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

   272

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

   273

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

   274

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

   275

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

   276

 * Once the language being parsed has been identified, 

   277

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

   278

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

   279

 *

   280

 * Aditionally, in sfc state it may further recursively enter the body state

   281

 * once again. This is because an sfc body may contain ACTIONS, which are then

   282

 * written in one of the three languages (ST, IL or SFC), so once again we need

   283

 * to figure out which language the ACTION in the SFC was written in. We already

   284

 * ahve all that done in the body state, so we recursively transition to the body 

   285

 * state once again.

   286

 * Note that in this case, when coming out of the st/il state (whichever language

   287

 * the action was written in) the sfc state will become active again. This is done by

   288

 * pushing and poping the previously active state!

   289

 *

   290

 * The sfc_qualifier_state is required because when parsing actions within an

   291

 * sfc, we will be expecting action qualifiers (N, P, R, S, DS, SD, ...). In order

   292

 * to bison to work correctly, these qualifiers must be returned as tokens. However,

   293

 * these tokens are not reserved keywords, which means it should be possible to

   294

 * define variables/functions/FBs with any of these names (including 

   295

 * S and R which are special because they are also IL operators). So, when we are not

   296

 * expecting any action qualifiers, flex does not return these tokens, and is free

   297

 * to interpret them as previously defined variables/functions/... as the case may be.

   298

 *

   299

 * The state machine has 7 possible states (INITIAL, config, decl, body, st, il, sfc)

   300

 * Possible state changes are:

   301

 *   INITIAL -> goto(decl_state)

   302

 *               (when a FUNCTION, FUNCTION_BLOCK, or PROGRAM is found,

   303

 *                and followed by a VAR declaration)

   304

 *   INITIAL -> goto(body_state) 

   305

 *                (when a FUNCTION, FUNCTION_BLOCK, or PROGRAM is found,

   306

 *                 and _not_ followed by a VAR declaration)

   307

 *                (This transition is actually commented out, since the syntax

   308

 *                 does not allow the declaration of functions, FBs, or programs

   309

 *                 without any VAR declaration!)

   310

 *   INITIAL -> goto(config_state)

   311

 *                (when a CONFIGURATION is found)

   312

 *   decl_state    -> push(decl_state); goto(body_state)

   313

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

   314

 *   decl_state    -> push(decl_state); goto(sfc_state)

   315

 *                     (when it figures out it is parsing sfc language)

   316

 *   body_state    -> goto(st_state)

   317

 *                     (when it figures out it is parsing st language)

   318

 *   body_state    -> goto(il_state)

   319

 *                     (when it figures out it is parsing il language)

   320

 *   st_state      -> pop()

   321

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   322

 *                      END_ACTION or END_TRANSITION is found)

   323

 *   il_state      -> pop()

   324

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   325

 *                      END_ACTION or END_TRANSITION is found)

   326

 *   decl_state    -> goto(INITIAL)

   327

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, or END_PROGRAM is found)

   328

 *   sfc_state     -> goto(INITIAL)

   329

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, or END_PROGRAM is found)

   330

 *   config_state  -> goto(INITIAL)

   331

 *                     (when a END_CONFIGURATION is found)

   332

 *   sfc_state     -> push(sfc_state); goto(body_state)

   333

 *                     (when parsing an action. This transition is requested by bison)

   334

 *   sfc_state     -> push(sfc_state); goto(sfc_qualifier_state)

   335

 *                     (when expecting an action qualifier. This transition is requested by bison)

   336

 *   sfc_qualifier_state -> pop()

   337

 *                     (when no longer expecting an action qualifier. This transition is requested by bison)

   338

 *

   339

 */

   340

   341

   342

   343

/* we are parsing a configuration. */

   344

%s config_state

   345

   346

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

   347

%s decl_state

   348

   349

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

   350

%x body_state

   351

   352

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

   353

%s il_state

   354

   355

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

   356

%s st_state

   357

   358

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

   359

%s sfc_state

   360

   361

/* we are parsing sfc code, and expecting an action qualifier.      */

   362

%s sfc_qualifier_state

   363

   364

   365

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

   366

/* File #include's */

   367

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

   368

   369

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

   370

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

   371

 * The accepted syntax is:

   372

 *  {#include "<filename>"}

   373

 */

   374

   375

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

   376

%x include_beg

   377

%x include_filename

   378

%x include_end

   379

   380

   381

file_include_pragma_filename	[^\"]*

   382

file_include_pragma_beg		"{#include"{st_whitespace_only}\"

   383

file_include_pragma_end		\"{st_whitespace_only}"}"

   384

file_include_pragma			{file_include_pragma_beg}{file_include_pragma_filename}{file_include_pragma_end}

   385

   386

   387

%{

   388

#define MAX_INCLUDE_DEPTH 16

   389

   390

typedef struct {

   391

	  YY_BUFFER_STATE buffer_state;

   392

	  int lineno;

   393

	  const char *filename;

   394

	} include_stack_t;

   395

   396

include_stack_t include_stack[MAX_INCLUDE_DEPTH];

   397

int include_stack_ptr = 0;

   398

   399

const char *INCLUDE_DIRECTORIES[] = {

   400

	DEFAULT_LIBDIR,

   401

	".",

   402

	"/lib",

   403

	"/usr/lib",

   404

	"/usr/lib/iec",

   405

	NULL /* must end with NULL!! */

   406

	};

   407

   408

%}

   409

   410

   411

   412

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

   413

/* Prelimenary constructs... */

   414

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

   415

   416

   417

/* A pragma... */

   418

   419

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

   420

   421

/* NOTE: this seemingly unnecessary complex definition is required

   422

 *       to be able to eat up comments such as:

   423

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

   424

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

   425

 *       since {comment} itself will later be used with

   426

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

   427

 */

   428

not_asterisk				[^*]

   429

not_close_parenthesis_nor_asterisk	[^*)]

   430

asterisk				"*"

   431

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

   432

   433

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

   434

   435

   436

/*

   437

3.1 Whitespace

   438

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

   439

 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.).

   440

 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.

   441

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

   442

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

   443

 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.).

   444

 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.

   445

 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.

   446

 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.

   447

*/

   448

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

   449

 *

   450

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

   451

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

   452

 *       for each language...

   453

 */

   454

/*

   455

 * NOTE: we cannot use

   456

 *         st_whitespace	[:space:]*

   457

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

   458

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

   459

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

   460

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

   461

 *       generate a "dangerous trailing context" warning!

   462

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

   463

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

   464

 */

   465

   466

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

   467

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

   468

   469

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

   470

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

   471

   472

st_whitespace	{st_whitespace_text}*

   473

il_whitespace	{il_whitespace_text}*

   474

   475

st_whitespace_text_no_pragma	{st_whitespace_only}|{comment}

   476

il_whitespace_text_no_pragma	{il_whitespace_only}|{comment}

   477

   478

st_whitespace_no_pragma	{st_whitespace_text_no_pragma}*

   479

il_whitespace_no_pragma	{il_whitespace_text_no_pragma}*

   480

   481

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

   482

   483

   484

   485

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

   486

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

   487

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

   488

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

   489

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

   490

 *       contains non-alphabetic characters!

   491

 *       The correct way of doing it would be to use

   492

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

   493

 *

   494

 *       Unfortunately, further on we need all printable

   495

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

   496

 *       Flex does not allow sets to be composed by excluding

   497

 *       elements. Sets may only be constructed by adding new

   498

 *       elements, which means that we have to revert to

   499

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

   500

 *       of the printable characters with the required exceptions.

   501

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

   502

 *       no way to work around it|

   503

 *

   504

 *       The conclusion is that our parser is limited to ASCII

   505

 *       based host computers!!

   506

 */

   507

letter		[A-Za-z]

   508

digit		[0-9]

   509

octal_digit	[0-7]

   510

hex_digit	{digit}|[A-F]

   511

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

   512

   513

   514

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

   515

/* B.1.2 Constants */

   516

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

   517

   518

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

   519

/* B.1.2.1   Numeric literals */

   520

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

   521

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