ledger-cli/doc/grammar.y

/**
 * @file grammar.y
 * @version 3.0
 * @author John Wiegley
 *
 * @brief Canonical BNF grammar for Ledger data files
 *
 * Extensions are permitted if: they are not required, and they are
 * backwards-compatible with this grammar.
 */

/*
 * There are three special terminals in this grammar, which violate its
 * context free nature:
 *
 * TEXT       -- consumes all characters until the next terminal
 *               or EOL (end of line)
 * WHITESPACE -- any amount of whitespace, not including EOL
 * STRING     -- characters up to the next WHITESPACE or EOL
 *
 * BIGINT     -- a number of any width, matching [0-9]+
 * INT4       -- a four digit wide number
 * INT2       -- a two digit wide number
 * INT1       -- a one digit wide number
 *
 * Except for 1) the 'spacer' production (see below), 2) EOL, and 3) the
 * WHITESPACE required to begin a posting, whitespace is otherwise
 * ignored.
 *
 * Yes, this grammar is confusing and not so happy for machine readers,
 * but it was designed for the human author and reader.  Once parsed,
 * the contents must be unambiguous, which means they can be output to
 * more rigorous formats for other programs to consume.
 */

/*
 * Journals
 *
 * A journal is a file which primarily contains xacts, among other elements.
 */

journal:
    journal_item journal |
    /* epsilon */
    ;

journal_item:
    whitespace
    directive |
    xact |
    ;

whitespace:
    EOL |
    WHITESPACE EOL |
    ';' TEXT EOL |              /* these next four are all ignored */
    '*' TEXT EOL |
    ;

directive:
    '@' word_directive EOL |
    '!' word_directive EOL |
    word_directive EOL |
    char_directive EOL
    ;

word_directive:
    "include" TEXT |
    "account" TEXT |
    "end" |
    "alias" STRING '=' TEXT |
    "def" TEXT |
    TEXT WHITESPACE TEXT        /* looked up in session (aka maybe Python) */
    ;

char_directive:
    'i' date time TEXT |        /* a timeclock.el "check in" */
    'I' date time TEXT |
    'o' date time TEXT |        /* a timeclock.el "check out" */
    'O' date time TEXT |
    'h' TEXT EOL |
    'b' TEXT EOL |
    'D' amount |                /* sets display parameters for a commodity */
    'A' TEXT |                  /* sets the "default balancing account" */
    'C' commodity '=' amount |  /* specifies a commodity conversion */
    'P' date time commodity amount | /* a pricing history xact */
    'N' commodity |             /* commodity's price is never downloaded */
    'Y' INT4 |                  /* sets the default year for date parsing */
    '-' '-' STRING TEXT |       /* specify command-line options in the file */
    ;

date: INT4 date_sep INT2 date_sep INT2 ;
date_opt: '=' date | /* epsilon */ ;
date_sep: '/' | '-' | '.' ;

time: INT2 ':' INT2 ':' INT2 ;

commodity:
    '"' TEXT '"' |
    STRING ;

/*
 * Xacts
 *
 * Xacts are the atomic units of accounting, which are composed of
 * multiple postings between accounts, so long as it all balances in
 * the end.
 */

xact: plain_xact |
       periodic_xact |
       automated_xact ;

plain_xact:
    date date_opt status_opt code_opt FULLSTRING note_opt EOL
    postings ;

status_opt: status | /* epsilon */ ;
status: '*' | '!' | /* epsilon */ ;

code_opt: code | /* epsilon */ ;
code: '(' TEXT ')' ;

spacer: ' ' ' ' | '\t' | ' ' '\t' ;

note_opt: spacer note | /* epsilon */ ;
note: ';' TEXT ;

/* ---------------------------------------------------------------------- */

periodic_xact:
    '~' period_expr note_opt EOL
    posting postings ;

/*
 * A period expression has its own sub-grammar, which I don't quite have
 * the time to exhaustively describe now.  See datetime.cc.  It allows
 * for lots and lots of things, and is probably horribly ambiguous.
 */

period_expr: FULLSTRING ;

/* ---------------------------------------------------------------------- */

automated_xact:
    '=' value_expr note_opt EOL
    posting postings ;

/*
 * Value expressions are a algebraic math expressions very similar to
 * XPath (minus the path traversal items).  This grammar needs fleshing
 * out also, since it's allowed in many places.
 */

value_expr: FULLSTRING ;

/*
 * There is a serious ambiguity here which the parser resolves as
 * follows: if an amount_expr can be parsed as an amount, it's an
 * amount; otherwise, it's a value expression.
 */

quantity: neg_opt BIGINT decimal_opt ;

neg_opt: '-' | /* epsilon */ ;
decimal_opt: '.' BIGINT | /* epsilon */ ;

annotation: lot_price_opt lot_date_opt lot_note_opt ;

lot_date_opt: date | /* epsilon */ ;
lot_date: '[' date ']' ;

lot_price_opt: price | /* epsilon */ ;
lot_price: '{' amount '}' ;

lot_note_opt: note | /* epsilon */ ;
lot_note: '(' string ')' ;

amount:
    neg_opt commodity quantity annotation |
    quantity commodity annotation ;

amount_expr: amount | value_expr ;

/*
 * Postings
 *
 * Postings are the fundamental unit of accounting, and represent
 * the movement of commodities to or from an account.  Thus, paying off
 * your credit card consists of two balancing postings: one that
 * withdraws money from your checking account, and another which pays
 * money to your credit institution.
 */

postings:
    posting postings |
    /* epsilon */
    ;

posting:
    WHITESPACE status_opt account values_opt note_opt EOL;

account_name: FULLSTRING ;

values_opt:
    spacer amount_expr price_opt |
    /* epsilon */
    ;

price_opt: price | /* epsilon */ ;
price:
    '@' amount_expr |
    '@@' amount_expr            /* in this case, it's the whole price */
    ;

account:
    account_name |
    '(' account_name ')' |
    '[' account_name ']' ;

/* grammar.y ends here */