1

/*

     2

 *  matiec - a compiler for the programming languages defined in IEC 61131-3

     3

 *

     4

 *  Copyright (C) 2003-2011  Mario de Sousa (msousa@fe.up.pt)

     5

 *

     6

 *  This program is free software: you can redistribute it and/or modify

     7

 *  it under the terms of the GNU General Public License as published by

     8

 *  the Free Software Foundation, either version 3 of the License, or

     9

 *  (at your option) any later version.

    10

 *

    11

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

    12

 *  but WITHOUT ANY WARRANTY; without even the implied warranty of

    13

 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

    14

 *  GNU General Public License for more details.

    15

 *

    16

 *  You should have received a copy of the GNU General Public License

    17

 *  along with this program.  If not, see <http://www.gnu.org/licenses/>.

    18

 *

    19

 *

    20

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

    21

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

    22

 */

    23

    24

/*

    25

 * An IEC 61131-3 compiler.

    26

 *

    27

 * Based on the

    28

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

    29

 *

    30

 */

    31

    32

/*

    33

 * Stage 1

    34

 * =======

    35

 *

    36

 * This file contains the lexical tokens definitions, from which

    37

 * the flex utility will generate a lexical parser function.

    38

 */

    39

    40

    41

    42

    43

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

    44

/* Lexical Parser Options... */

    45

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

    46

    47

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

    48

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

    49

 * programs being written inter-actively by the user.

    50

 * This option saves the resulting parser from calling the

    51

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

    52

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

    53

 */

    54

%option never-interactive

    55

    56

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

    57

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

    58

 */

    59

%pointer

    60

    61

    62

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

    63

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

    64

 * the resulting text handed up to the syntax parser

    65

 * will not be changed, and keep the original case

    66

 * of the letters in the input file.

    67

 */

    68

%option case-insensitive

    69

    70

/* Have the generated lexical analyser keep track of the

    71

 * line number it is currently analysing.

    72

 * This is used to pass up to the syntax parser

    73

 * the number of the line on which the current

    74

 * token was found. It will enable the syntax parser

    75

 * to generate more informatve error messages...

    76

 */

    77

%option yylineno

    78

    79

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

    80

 * yy_push_state() functions

    81

 */

    82

%option stack

    83

    84

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

    85

 * the yy_top_state(), however this function is not

    86

 * currently being used. This means that the compiler

    87

 * is complaining about the existance of this function.

    88

 * The following option removes the yy_top_state()

    89

 * function from the resulting c code, so the compiler 

    90

 * no longer complains.

    91

 */

    92

%option noyy_top_state

    93

    94

/* We will be using unput() in our flex code, so we cannot set the following option!... */

    95

/*

    96

%option nounput

    97

*/

    98

    99

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

   100

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

   101

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

   102

   103

   104

%{

   105

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

   106

 * Useful for testing the parser generated by flex.

   107

 */

   108

/*

   109

#define TEST_MAIN

   110

*/

   111

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

   112

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

   113

 * in the syntax parser header file...

   114

 */

   115

#ifdef TEST_MAIN

   116

#define DEFAULT_LIBDIR "just_testing"

   117

#endif

   118

   119

   120

   121

/* Required for strdup() */

   122

#include <string.h>

   123

   124

/* Required only for the declaration of abstract syntax classes

   125

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

   126

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

   127

 * in iec_bison.h contains pointers to these classes, so we must include

   128

 * it here.

   129

 */

   130

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

   131

   132

   133

/* iec_bison.h is generated by bison.

   134

 * Contains the definition of the token constants, and the

   135

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

   136

 */

   137

#include "iec_bison.h"

   138

#include "stage1_2_priv.hh"

   139

   140

   141

/* Variable defined by the bison parser,

   142

 * where the value of the tokens will be stored

   143

 */

   144

extern YYSTYPE yylval;

   145

   146

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

   147

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

   148

 * Note that flex accesses and updates this global variable

   149

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

   150

 * directive...

   151

 */

   152

/*

   153

 NOTE: already defined in iec_bison.h 

   154

extern const char *current_filename;

   155

*/

   156

   157

   158

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

   159

/* NOTE: it seems that this #define is no longer needed, It has been 

   160

 * replaced by %option nounput.

   161

 * Should we simply delete it?

   162

 * For now leave it in, in case someone is using an old version of flex.

   163

 * In any case, the most harm that can result in a warning message

   164

 * when compiling iec.flex.c:

   165

 * warning: ‘void yyunput(int, char*)’ defined but not used

   166

 */

   167

#define YY_NO_UNPUT

   168

   169

/* Variable defined by the bison parser.

   170

 * It must be initialised with the location

   171

 * of the token being parsed.

   172

 * This is only needed if we want to keep

   173

 * track of the locations, in order to give

   174

 * more meaningful error messages!

   175

 */

   176

/*

   177

 *extern YYLTYPE yylloc;

   178

b*/

   179

#define YY_INPUT(buf,result,max_size)  {\

   180

    result = GetNextChar(buf, max_size);\

   181

    if (  result <= 0  )\

   182

      result = YY_NULL;\

   183

    }

   184

   185

   186

/* A counter to track the order by which each token is processed.

   187

 * NOTE: This counter is not exactly linear (i.e., it does not get incremented by 1 for each token).

   188

 *       i.e.. it may get incremented by more than one between two consecutive tokens.

   189

 *       This is due to the fact that the counter gets incremented every 'user action' in flex,

   190

 *       however not every user action will result in a token being passed to bison.

   191

 *       Nevertheless this is still OK, as we are only interested in the relative

   192

 *       ordering of tokens...

   193

 */

   194

static long int current_order = 0;

   195

   196

   197

/* Macro that is executed for every action.

   198

 * We use it to pass the location of the token

   199

 * back to the bison parser...

   200

 */

   201

#define YY_USER_ACTION {\

   202

	yylloc.first_line = current_tracking->lineNumber;			\

   203

	yylloc.first_column = current_tracking->currentTokenStart;		\

   204

	yylloc.first_file = current_filename;					\

   205

	yylloc.first_order = current_order;					\

   206

	yylloc.last_line = current_tracking->lineNumber;			\

   207

	yylloc.last_column = current_tracking->currentChar - 1;			\

   208

	yylloc.last_file = current_filename;					\

   209

	yylloc.last_order = current_order;					\

   210

	current_tracking->currentTokenStart = current_tracking->currentChar;	\

   211

	current_order++;							\

   212

	}

   213

   214

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

   215

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

   216

 * one...

   217

 * Lets check a few random characters...

   218

 */

   219

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

   220

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

   221

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

   222

#endif

   223

   224

   225

/* Function only called from within flex, but defined

   226

 * in iec.y!

   227

 * We declare it here...

   228

 *

   229

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

   230

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

   231

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

   232

 * does it continue searching in the library elements

   233

 */

   234

//token_id_t get_identifier_token(const char *identifier_str);

   235

int get_identifier_token(const char *identifier_str);

   236

%}

   237

   238

   239

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

   240

/* Forward Declaration of functions defined later. */

   241

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

   242

   243

%{

   244

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

   245

void unput_text(unsigned int n);

   246

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

   247

 * but first return to the stream an additional character to mark the end of the token. 

   248

 */

   249

void unput_and_mark(const char c);

   250

%}

   251

   252

   253

   254

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

   255

/* Lexical Parser States... */

   256

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

   257

   258

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

   259

 *       within the same file.

   260

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

   261

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

   262

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

   263

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

   264

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

   265

 *

   266

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

   267

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

   268

 *            token to simply ignore it. This makes the syntax

   269

 *            definition (in iec.y) very cluttered!

   270

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

   271

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

   272

 *            the EOL token. This requires the lexical parser

   273

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

   274

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

   275

 *            very difficult to understand the lexical parser as it

   276

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

   277

 *            between the states. The state transitions would be

   278

 *            intermingled with the lexical parser defintion!

   279

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

   280

 *            merely distinguishes between function headers and function

   281

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

   282

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

   283

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

   284

 *            whitespace.

   285

 *            The ST language syntax has been augmented in the syntax

   286

 *            parser configuration to ignore any EOL tokens that it may

   287

 *            come across!

   288

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

   289

 *            much less intensely.

   290

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

   291

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

   292

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

   293

 *            the lexical parser becomes very simple. All state transitions

   294

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

   295

 *            easily identified. Thus knowledge of the syntax required by

   296

 *            the lexical parser is very limited!

   297

 *

   298

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

   299

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

   300

 * The lexical parser will enter the body state when

   301

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

   302

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

   303

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

   304

 * lexical analyser with this option!

   305

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

   306

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

   307

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

   308

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

   309

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

   310

 * could no longer be resolved.

   311

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

   312

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

   313

 * This option was therefore dropped in favour of another!

   314

 *

   315

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

   316

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

   317

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

   318

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

   319

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

   320

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

   321

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

   322

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

   323

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

   324

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

   325

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

   326

 * Once the language being parsed has been identified, 

   327

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

   328

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

   329

 *

   330

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

   331

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

   332

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

   333

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

   334

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

   335

 * state once again.

   336

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

   337

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

   338

 * pushing and poping the previously active state!

   339

 *

   340

 * The sfc_qualifier_state is required because when parsing actions within an

   341

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

   342

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

   343

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

   344

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

   345

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

   346

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

   347

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

   348

 *

   349

 * The time_literal_state is required because TIME# literals are decomposed into 

   350

 * portions, and wewant to send these portions one by one to bison. Each poertion will 

   351

 * represent the value in days/hours/minutes/seconds/ms.

   352

 * Unfortunately, some of these portions may also be lexically analysed as an identifier. So,

   353

 * we need to disable lexical identification of identifiers while parsing TIME# literals!

   354

 * e.g.:  TIME#55d_4h_56m

   355

 *       We would like to return to bison the tokens 'TIME' '#' '55d' '_' '4h' '_' '56m'

   356

 *       Unfortunately, flex will join '_' and '4h' to create a legal {identifier} '_4h',

   357

 *       and return that identifier instead! So, we added this state!

   358

 *

   359

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

   360

 * Possible state changes are:

   361

 *   INITIAL -> goto(decl_state)

   362

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

   363

 *                and followed by a VAR declaration)

   364

 *   INITIAL -> goto(body_state) 

   365

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

   366

 *                 and _not_ followed by a VAR declaration)

   367

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

   368

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

   369

 *                 without any VAR declaration!)

   370

 *   INITIAL -> goto(config_state)

   371

 *                (when a CONFIGURATION is found)

   372

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

   373

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

   374

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

   375

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

   376

 *   body_state    -> goto(st_state)

   377

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

   378

 *   body_state    -> goto(il_state)

   379

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

   380

 *   st_state      -> pop()

   381

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   382

 *                      END_ACTION or END_TRANSITION is found)

   383

 *   il_state      -> pop()

   384

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   385

 *                      END_ACTION or END_TRANSITION is found)

   386

 *   decl_state    -> goto(INITIAL)

   387

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

   388

 *   sfc_state     -> goto(INITIAL)

   389

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

   390

 *   config_state  -> goto(INITIAL)

   391

 *                     (when a END_CONFIGURATION is found)

   392

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

   393

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

   394

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

   395

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

   396

 *   sfc_qualifier_state -> pop()

   397

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

   398

 *   config_state  -> push(config_state); goto(task_init_state)

   399

 *                     (when parsing a task initialisation. This transition is requested by bison)

   400

 *   task_init_state -> pop()

   401

 *                     (when no longer parsing task initialisation parameters. This transition is requested by bison)

   402

 *

   403

 */

   404

   405

   406

/* we are parsing a configuration. */

   407

%s config_state

   408

   409

/* Inside a configuration, we are parsing a task initialisation parameters */

   410

/* This means that PRIORITY, SINGLE and INTERVAL must be handled as

   411

 * tokens, and not as possible identifiers. Note that the above words

   412

 * are not keywords.

   413

 */

   414

%s task_init_state

   415

   416

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

   417

%s decl_state

   418

   419

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

   420

%x body_state

   421

   422

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

   423

%s il_state

   424

   425

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

   426

%s st_state

   427

   428

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

   429

%s sfc_state

   430

   431

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

   432

%s sfc_qualifier_state

   433

   434

/* we are parsing sfc code, and expecting the priority token.       */

   435

%s sfc_priority_state

   436

   437

/* we are parsing a TIME# literal. We must not return any {identifier} tokens. */

   438

%x time_literal_state

   439

   440

   441

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

   442

/* File #include's */

   443

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

   444

   445

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

   446

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

   447

 * The accepted syntax is:

   448

 *  {#include "<filename>"}

   449

 */

   450

   451

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

   452

%x include_beg

   453

%x include_filename

   454

%x include_end

   455

   456

   457

file_include_pragma_filename	[^\"]*

   458

file_include_pragma_beg		"{#include"{st_whitespace_only}\"

   459

file_include_pragma_end		\"{st_whitespace_only}"}"

   460

file_include_pragma			{file_include_pragma_beg}{file_include_pragma_filename}{file_include_pragma_end}

   461

   462

   463

%{

   464

#define MAX_INCLUDE_DEPTH 16

   465

   466

typedef struct {

   467

	  YY_BUFFER_STATE buffer_state;

   468

	  tracking_t* env;

   469

	  const char *filename;

   470

	} include_stack_t;

   471

   472

tracking_t* current_tracking;

   473

include_stack_t include_stack[MAX_INCLUDE_DEPTH];

   474

int include_stack_ptr = 0;

   475

   476

const char *INCLUDE_DIRECTORIES[] = {

   477

	DEFAULT_LIBDIR,

   478

	".",

   479

	"/lib",

   480

	"/usr/lib",

   481

	"/usr/lib/iec",

   482

	NULL /* must end with NULL!! */

   483

	};

   484

   485

%}

   486

   487

   488

   489

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

   490

/* Prelimenary constructs... */

   491

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

   492

   493

/* In order to allow the declaration of POU prototypes (Function, FB, Program, ...),

   494

 * especially the prototypes of Functions and FBs defined in the standard

   495

 * (i.e. standard functions and FBs), we extend the IEC 61131-3 standard syntax 

   496

 * with two pragmas to indicate that the code is to be parsed (going through the 

   497

 * lexical, syntactical, and semantic analysers), but no code is to be generated.

   498

 * 

   499

 * The accepted syntax is:

   500

 *  {no_code_generation begin}

   501

 *    ... prototypes ...

   502

 *  {no_code_generation end}

   503

 * 

   504

 * When parsing these prototypes the abstract syntax tree will be populated as usual,

   505

 * allowing the semantic analyser to correctly analyse the semantics of calls to these

   506

 * functions/FBs. However, stage4 will simply ignore all IEC61131-3 code

   507

 * between the above two pragmas.

   508

 */

   509

   510

disable_code_generation_pragma	"{disable code generation}"

   511