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

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

   159

    result = GetNextChar(buf, max_size);\

   160

    if (  result <= 0  )\

   161

      result = YY_NULL;\

   162

    }

   163

   164

/* Macro that is executed for every action.

   165

 * We use it to pass the location of the token

   166

 * back to the bison parser...

   167

 */

   168

#define YY_USER_ACTION {\

   169

	  yylloc.first_line = current_tracking->lineNumber;\

   170

  	yylloc.first_column = current_tracking->currentTokenStart;\

   171

  	yylloc.last_line = current_tracking->lineNumber;\

   172

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

   173

  	current_tracking->currentTokenStart = current_tracking->currentChar;\

   174

	}

   175

   176

   177

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

   178

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

   179

 * one...

   180

 * Lets check a few random characters...

   181

 */

   182

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

   183

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

   184

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

   185

#endif

   186

   187

   188

/* Function only called from within flex, but defined

   189

 * in iec.y!

   190

 * We declare it here...

   191

 *

   192

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

   193

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

   194

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

   195

 * does it continue searching in the library elements

   196

 */

   197

//token_id_t get_identifier_token(const char *identifier_str);

   198

int get_identifier_token(const char *identifier_str);

   199

%}

   200

   201

   202

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

   203

/* Forward Declaration of functions defined later. */

   204

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

   205

   206

%{

   207

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

   208

void unput_text(unsigned int n);

   209

%}

   210

   211

   212

   213

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

   214

/* Lexical Parser States... */

   215

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

   216

   217

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

   218

 *       within the same file.

   219

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

   220

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

   221

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

   222

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

   223

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

   224

 *

   225

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

   226

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

   227

 *            token to simply ignore it. This makes the syntax

   228

 *            definition (in iec.y) very cluttered!

   229

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

   230

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

   231

 *            the EOL token. This requires the lexical parser

   232

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

   233

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

   234

 *            very difficult to understand the lexical parser as it

   235

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

   236

 *            between the states. The state transitions would be

   237

 *            intermingled with the lexical parser defintion!

   238

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

   239

 *            merely distinguishes between function headers and function

   240

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

   241

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

   242

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

   243

 *            whitespace.

   244

 *            The ST language syntax has been augmented in the syntax

   245

 *            parser configuration to ignore any EOL tokens that it may

   246

 *            come across!

   247

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

   248

 *            much less intensely.

   249

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

   250

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

   251

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

   252

 *            the lexical parser becomes very simple. All state transitions

   253

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

   254

 *            easily identified. Thus knowledge of the syntax required by

   255

 *            the lexical parser is very limited!

   256

 *

   257

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

   258

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

   259

 * The lexical parser will enter the body state when

   260

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

   261

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

   262

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

   263

 * lexical analyser with this option!

   264

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

   265

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

   266

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

   267

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

   268

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

   269

 * could no longer be resolved.

   270

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

   271

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

   272

 * This option was therefore dropped in favour of another!

   273

 *

   274

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

   275

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

   276

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

   277

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

   278

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

   279

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

   280

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

   281

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

   282

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

   283

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

   284

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

   285

 * Once the language being parsed has been identified, 

   286

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

   287

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

   288

 *

   289

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

   290

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

   291

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

   292

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

   293

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

   294

 * state once again.

   295

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

   296

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

   297

 * pushing and poping the previously active state!

   298

 *

   299

 * The sfc_qualifier_state is required because when parsing actions within an

   300

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

   301

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

   302

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

   303

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

   304

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

   305

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

   306

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

   307

 *

   308

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

   309

 * Possible state changes are:

   310

 *   INITIAL -> goto(decl_state)

   311

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

   312

 *                and followed by a VAR declaration)

   313

 *   INITIAL -> goto(body_state) 

   314

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

   315

 *                 and _not_ followed by a VAR declaration)

   316

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

   317

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

   318

 *                 without any VAR declaration!)

   319

 *   INITIAL -> goto(config_state)

   320

 *                (when a CONFIGURATION is found)

   321

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

   322

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

   323

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

   324

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

   325

 *   body_state    -> goto(st_state)

   326

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

   327

 *   body_state    -> goto(il_state)

   328

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

   329

 *   st_state      -> pop()

   330

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   331

 *                      END_ACTION or END_TRANSITION is found)

   332

 *   il_state      -> pop()

   333

 *                     (when a END_FUNCTION, END_FUNCTION_BLOCK, END_PROGRAM,

   334

 *                      END_ACTION or END_TRANSITION is found)

   335

 *   decl_state    -> goto(INITIAL)

   336

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

   337

 *   sfc_state     -> goto(INITIAL)

   338

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

   339

 *   config_state  -> goto(INITIAL)

   340

 *                     (when a END_CONFIGURATION is found)

   341

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

   342

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

   343

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

   344

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

   345

 *   sfc_qualifier_state -> pop()

   346

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

   347

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

   348

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

   349

 *   task_init_state -> pop()

   350

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

   351

 *

   352

 */

   353

   354

   355

/* we are parsing a configuration. */

   356

%s config_state

   357

   358

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

   359

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

   360

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

   361

 * are not keywords.

   362

 */

   363

%s task_init_state

   364

   365

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

   366

%s decl_state

   367

   368

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

   369

%x body_state

   370

   371

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

   372

%s il_state

   373

   374

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

   375

%s st_state

   376

   377

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

   378

%s sfc_state

   379

   380

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

   381

%s sfc_qualifier_state

   382

   383

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

   384

%s sfc_priority_state

   385

   386

   387

   388

   389

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

   390

/* File #include's */

   391

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

   392

   393

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

   394

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

   395

 * The accepted syntax is:

   396

 *  {#include "<filename>"}

   397

 */

   398

   399

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

   400

%x include_beg

   401

%x include_filename

   402

%x include_end

   403

   404

   405

file_include_pragma_filename	[^\"]*

   406

file_include_pragma_beg		"{#include"{st_whitespace_only}\"

   407

file_include_pragma_end		\"{st_whitespace_only}"}"

   408

file_include_pragma			{file_include_pragma_beg}{file_include_pragma_filename}{file_include_pragma_end}

   409

   410

   411

%{

   412

#define MAX_INCLUDE_DEPTH 16

   413

   414

typedef struct {

   415

	  YY_BUFFER_STATE buffer_state;

   416

	  tracking_t* env;

   417

	  const char *filename;

   418

	} include_stack_t;

   419

   420

tracking_t* current_tracking;

   421

include_stack_t include_stack[MAX_INCLUDE_DEPTH];

   422

int include_stack_ptr = 0;

   423

   424

const char *INCLUDE_DIRECTORIES[] = {

   425

	DEFAULT_LIBDIR,

   426

	".",

   427

	"/lib",

   428

	"/usr/lib",

   429

	"/usr/lib/iec",

   430

	NULL /* must end with NULL!! */

   431

	};

   432

   433

%}

   434

   435

   436

   437

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

   438

/* Prelimenary constructs... */

   439

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

   440

   441

   442

/* A pragma... */

   443

   444

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

   445

   446

/* NOTE: this seemingly unnecessary complex definition is required

   447

 *       to be able to eat up comments such as:

   448

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

   449

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

   450

 *       since {comment} itself will later be used with

   451

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

   452

 */

   453

not_asterisk				[^*]

   454

not_close_parenthesis_nor_asterisk	[^*)]

   455

asterisk				"*"

   456

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

   457

   458

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

   459

   460

   461

/*

   462

3.1 Whitespace

   463

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

   464

 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.).

   465

 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.

   466

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

   467

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

   468

 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.).

   469

 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.

   470

 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.

   471

 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.

   472

*/

   473

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

   474

 *

   475

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

   476

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

   477

 *       for each language...

   478

 */

   479

/*

   480

 * NOTE: we cannot use

   481

 *         st_whitespace	[:space:]*

   482

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

   483

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

   484

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

   485

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

   486

 *       generate a "dangerous trailing context" warning!

   487

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

   488

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

   489

 */

   490

   491

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

   492

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

   493

   494

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

   495

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

   496

   497

st_whitespace	{st_whitespace_text}*

   498

il_whitespace	{il_whitespace_text}*

   499

   500

st_whitespace_text_no_pragma	{st_whitespace_only}|{comment}

   501

il_whitespace_text_no_pragma	{il_whitespace_only}|{comment}

   502

   503

st_whitespace_no_pragma	{st_whitespace_text_no_pragma}*

   504

il_whitespace_no_pragma	{il_whitespace_text_no_pragma}*

   505

   506

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

   507

   508

   509

   510

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

   511

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

   512

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

   513

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

   514

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

   515

 *       contains non-alphabetic characters!

   516

 *       The correct way of doing it would be to use

   517

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

   518

 *