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

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

    95

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

    96

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

    97

    98

    99

%{

   100

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

   101

 * Useful for testing the parser generated by flex.

   102

 */

   103

/*

   104

#define TEST_MAIN

   105

*/

   106

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

   107

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

   108

 * in the syntax parser header file...

   109

 */

   110

#ifdef TEST_MAIN

   111

#define DEFAULT_LIBDIR "just_testing"

   112

#endif

   113

   114

   115

   116

/* Required for strdup() */

   117

#include <string.h>

   118

   119

/* Required only for the declaration of abstract syntax classes

   120

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

   121

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

   122

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

   123

 * it here.

   124

 */

   125

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

   126

   127

   128

/* generated by bison.

   129

 * Contains the definition of the token constants, and the

   130

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

   131

 */

   132

#include "iec.y.hh"

   133

#include "stage1_2_priv.hh"

   134

   135

   136

/* Variable defined by the bison parser,

   137

 * where the value of the tokens will be stored

   138

 */

   139

extern YYSTYPE yylval;

   140

   141

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

   142

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

   143

 * Note that flex accesses and updates this global variable

   144

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

   145

 * directive...

   146

 */

   147

/*

   148

 NOTE: already defined in iec.y.hh 

   149

extern const char *current_filename;

   150

*/

   151

   152

   153

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

   154

#define YY_NO_UNPUT

   155

   156

/* Variable defined by the bison parser.

   157

 * It must be initialised with the location

   158

 * of the token being parsed.

   159

 * This is only needed if we want to keep

   160

 * track of the locations, in order to give

   161

 * more meaningful error messages!

   162

 */

   163

extern YYLTYPE yylloc;

   164

   165

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

   166

    result = GetNextChar(buf, max_size);\

   167

    if (  result <= 0  )\

   168

      result = YY_NULL;\

   169

    }

   170

   171

/* Macro that is executed for every action.

   172

 * We use it to pass the location of the token

   173

 * back to the bison parser...

   174

 */

   175

#define YY_USER_ACTION {\

   176

	  yylloc.first_line = current_tracking->lineNumber;\

   177

  	yylloc.first_column = current_tracking->currentTokenStart;\

   178

  	yylloc.last_line = current_tracking->lineNumber;\

   179

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

   180

  	current_tracking->currentTokenStart = current_tracking->currentChar;\

   181

	}

   182

   183

   184

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

   185

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

   186

 * one...

   187

 * Lets check a few random characters...

   188

 */

   189

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

   190

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

   191

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

   192

#endif

   193

   194

   195

/* Function only called from within flex, but defined

   196

 * in iec.y!

   197

 * We declare it here...

   198

 *

   199

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

   200

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

   201

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

   202

 * does it continue searching in the library elements

   203

 */

   204

//token_id_t get_identifier_token(const char *identifier_str);

   205

int get_identifier_token(const char *identifier_str);

   206

%}

   207

   208

   209

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

   210

/* Forward Declaration of functions defined later. */

   211

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

   212

   213

%{

   214

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

   215

void unput_text(unsigned int n);

   216

%}

   217

   218

   219

   220

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

   221

/* Lexical Parser States... */

   222

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

   223

   224

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

   225

 *       within the same file.

   226

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

   227

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

   228

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

   229

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

   230

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

   231

 *

   232

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

   233

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

   234

 *            token to simply ignore it. This makes the syntax

   235

 *            definition (in iec.y) very cluttered!

   236

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

   237

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

   238

 *            the EOL token. This requires the lexical parser

   239

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

   240

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

   241

 *            very difficult to understand the lexical parser as it

   242

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

   243

 *            between the states. The state transitions would be

   244

 *            intermingled with the lexical parser defintion!

   245

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

   246

 *            merely distinguishes between function headers and function

   247

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

   248

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

   249

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

   250

 *            whitespace.

   251

 *            The ST language syntax has been augmented in the syntax

   252

 *            parser configuration to ignore any EOL tokens that it may

   253

 *            come across!

   254

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

   255

 *            much less intensely.

   256

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

   257

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

   258

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

   259

 *            the lexical parser becomes very simple. All state transitions

   260

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

   261

 *            easily identified. Thus knowledge of the syntax required by

   262

 *            the lexical parser is very limited!

   263

 *

   264

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

   265

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

   266

 * The lexical parser will enter the body state when

   267

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

   268

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

   269

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

   270

 * lexical analyser with this option!

   271

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

   272

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

   273

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

   274

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

   275

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

   276

 * could no longer be resolved.

   277

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

   278

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

   279

 * This option was therefore dropped in favour of another!

   280

 *

   281

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

   282

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

   283

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

   284

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

   285

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

   286

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

   287

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

   288

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

   289

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

   290

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

   291

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

   292

 * Once the language being parsed has been identified, 

   293

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

   294

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

   295

 *

   296

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

   297

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

   298

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

   299

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

   300

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

   301

 * state once again.

   302

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

   303

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

   304

 * pushing and poping the previously active state!

   305

 *

   306

 * The sfc_qualifier_state is required because when parsing actions within an

   307

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

   308

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

   309

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

   310

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

   311

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

   312

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

   313

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

   314

 *

   315

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

   316

 * Possible state changes are:

   317

 *   INITIAL -> goto(decl_state)

   318

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

   319

 *                and followed by a VAR declaration)

   320

 *   INITIAL -> goto(body_state) 

   321

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

   322

 *                 and _not_ followed by a VAR declaration)

   323

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

   324

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

   325

 *                 without any VAR declaration!)

   326

 *   INITIAL -> goto(config_state)

   327

 *                (when a CONFIGURATION is found)

   328

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

   329

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

   330

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

   331

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

   332

 *   body_state    -> goto(st_state)

   333

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

   334

 *   body_state    -> goto(il_state)

   335

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

   336

 *   st_state      -> pop()

   337

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   338

 *                      END_ACTION or END_TRANSITION is found)

   339

 *   il_state      -> pop()

   340

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   341

 *                      END_ACTION or END_TRANSITION is found)

   342

 *   decl_state    -> goto(INITIAL)

   343

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

   344

 *   sfc_state     -> goto(INITIAL)

   345

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

   346

 *   config_state  -> goto(INITIAL)

   347

 *                     (when a END_CONFIGURATION is found)

   348

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

   349

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

   350

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

   351

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

   352

 *   sfc_qualifier_state -> pop()

   353

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

   354

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

   355

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

   356

 *   task_init_state -> pop()

   357

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

   358

 *

   359

 */

   360

   361

   362

/* we are parsing a configuration. */

   363

%s config_state

   364

   365

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

   366

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

   367

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

   368

 * are not keywords.

   369

 */

   370

%s task_init_state

   371

   372

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

   373

%s decl_state

   374

   375

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

   376

%x body_state

   377

   378

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

   379

%s il_state

   380

   381

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

   382

%s st_state

   383

   384

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

   385

%s sfc_state

   386

   387

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

   388

%s sfc_qualifier_state

   389

   390

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

   391

%s sfc_priority_state

   392

   393

   394

   395

   396

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

   397

/* File #include's */

   398

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

   399

   400

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

   401

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

   402

 * The accepted syntax is:

   403

 *  {#include "<filename>"}

   404

 */

   405

   406

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

   407

%x include_beg

   408

%x include_filename

   409

%x include_end

   410

   411

   412

file_include_pragma_filename	[^\"]*

   413

file_include_pragma_beg		"{#include"{st_whitespace_only}\"

   414

file_include_pragma_end		\"{st_whitespace_only}"}"

   415

file_include_pragma			{file_include_pragma_beg}{file_include_pragma_filename}{file_include_pragma_end}

   416

   417

   418

%{

   419

#define MAX_INCLUDE_DEPTH 16

   420

   421

typedef struct {

   422

	  YY_BUFFER_STATE buffer_state;

   423

	  tracking_t* env;

   424

	  const char *filename;

   425

	} include_stack_t;

   426

   427

tracking_t* current_tracking;

   428

include_stack_t include_stack[MAX_INCLUDE_DEPTH];

   429

int include_stack_ptr = 0;

   430

   431

const char *INCLUDE_DIRECTORIES[] = {

   432

	DEFAULT_LIBDIR,

   433

	".",

   434

	"/lib",

   435

	"/usr/lib",

   436

	"/usr/lib/iec",

   437

	NULL /* must end with NULL!! */

   438

	};

   439

   440

%}

   441

   442

   443

   444

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

   445

/* Prelimenary constructs... */

   446

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

   447

   448

   449

/* A pragma... */

   450

   451

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

   452

   453

/* NOTE: this seemingly unnecessary complex definition is required

   454

 *       to be able to eat up comments such as:

   455

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

   456

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

   457

 *       since {comment} itself will later be used with

   458

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

   459

 */

   460

not_asterisk				[^*]

   461

not_close_parenthesis_nor_asterisk	[^*)]

   462

asterisk				"*"

   463

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

   464

   465

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

   466

   467

   468

/*

   469

3.1 Whitespace

   470

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

   471

 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.).

   472

 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.

   473

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

   474

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

   475

 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.).

   476

 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.

   477

 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.

   478

 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.

   479

*/

   480

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

   481

 *

   482

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

   483

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

   484

 *       for each language...

   485

 */

   486

/*

   487

 * NOTE: we cannot use

   488

 *         st_whitespace	[:space:]*

   489

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

   490

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

   491

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

   492

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

   493

 *       generate a "dangerous trailing context" warning!

   494

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

   495

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

   496

 */

   497

   498

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

   499

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

   500

   501

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

   502

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

   503

   504

st_whitespace	{st_whitespace_text}*

   505

il_whitespace	{il_whitespace_text}*

   506

   507

st_whitespace_text_no_pragma	{st_whitespace_only}|{comment}

   508

il_whitespace_text_no_pragma	{il_whitespace_only}|{comment}

   509

   510

st_whitespace_no_pragma	{st_whitespace_text_no_pragma}*

   511

il_whitespace_no_pragma	{il_whitespace_text_no_pragma}*

   512

   513

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

   514

   515

   516

   517