fbtxt-lexer 0.0.1

data/lib/fbtxt/lexer/token-goals.rb

@@ -0,0 +1,306 @@
+module Fbtxt
+class Lexer
+
+
+######################################################
+## goal mode
+##  note - must be enclosed in ()!!!
+##
+##          todo - add () in basics - why? why not?
+
+
+
+
+## note - assume lines starting with opening ( are goal lines!!!!
+##  note - use \A (instead of ^) - \A strictly matches the start of the string.
+##
+##   note -  check for negative lookahead
+##                 to exclude ord (numbers) e.g.  (1), (42), etc.!!!
+##
+##  todo/fix -- exclude (a), (h), (n)  - TEAM_AWAY, TEAM_HOME, TEAM_NEUTRAL tokens!!
+
+START_GOAL_LINE_RE = %r{
+                     \A
+                        [ ]*    ## ignore leading spaces (if any)
+                       \(
+
+                       # check NEGATIVE lookahead
+                       (?!
+                             ##  exclude (a), (h), (n)
+                             ##    TEAM_AWAY, TEAM_HOME, TEAM_NEUTRAL
+                             (?: a|h|n )
+                             \)
+                        )
+
+ }xi
+
+
+
+#############
+##  check for goal compat(ility) "legacy" line
+##         e.g.
+## (6' Puskás 0-1, 9' Czibor 0-2, 11' Morlock 1-2, 18' Rahn 2-2,
+##   84' Rahn 3-2)
+## (6 Puskás 0-1, 9 Czibor 0-2, 11 Morlock 1-2, 18 Rahn 2-2,
+##  84 Rahn 3-2)
+
+
+START_GOAL_LINE_COMPAT_RE = %r{
+                   \A
+                        [ ]*    ## ignore leading spaces (if any)
+                      \(
+
+                      ## (i) check NEGATIVE lookahead
+                      ##    exclude score e.g. 1-1 etc.
+                          (?! [ ]* \b \d-\d \b)
+
+                      ## (ii) check POSITIVE lookahead
+                          (?= [ ]*
+                               \d{1,3}
+                                   '?    ## optional minute marker
+                                  (?: \+
+                                      \d{1,2}
+                                    '?    ## optional minute marker
+                                  )?
+                            )
+}xi
+
+
+
+###
+##  check for goal line (alternate syntax)
+##    (1-0 Player, 1-1 Player, ...)
+#    must start-off OR yes, include score
+##
+##    note - allow "centered" style e.g.
+##           (    Player 44' (p)  1-0
+##                                1-1 Player 64'   )
+START_GOAL_LINE_ALT_RE = %r{
+                     \A
+                        [ ]*    ## ignore leading spaces (if any)
+                      \(
+
+                      # check POSITIVE lookahead
+                       (?=  .*?         ## note - non-greedy
+                                \b \d-\d \b    ## score e.g. 0-1
+                         )
+                 }xi
+
+
+
+
+
+
+
+###
+#  note - alternate goal separator dash (-) MUST have leading and trailing space!!!
+#    e.g.   (Metzger 83 - Krämer 29, 88, Cichy 33, Rahn 37)
+#    e.g.   (Metzger - Krämer (2), Cichy, Rahn)
+#            (Brunnenmeier 17 - Gerwien 74)
+#            (Brunnenmeier - Gerwien)
+#    that is,  NOT allowed
+#    e.g.   (Metzger 83-Krämer 29, 88, Cichy 33, Rahn 37)
+#            (Brunnenmeier 17-Gerwien 74)
+#            (Brunnenmeier-Gerwien)
+#
+#   note - allow split by - e.g.
+#     Frankfurt   4-2 Schalke     (Kreß 45, Solz 55, Trimhold 58, Huberts 73 p -
+#                                  Berz 7, Herrmann 74)
+
+
+GOAL_SEP_ALT_RE = %r{
+          (?<goal_sep_alt>
+              (?<=[ ])   ## positive lookbehind - space required
+              -
+              (?=[ ]|\z)    ## positive lookahead - speace required
+             )}x
+
+
+## e.g.  (2)
+##       (2/p), (2/pen.), (3/2p), (3/ 2 pen.)
+##      -or-  (2,1pen), (3, 2 pens)
+##
+##       (p), (pen.) (2 pen.), (2p)
+##       (og), (o.g.),
+##        (2og), (2 o.g.), (2ogs)
+#
+##
+
+GOAL_COUNT_RE = %r{
+   (?<goal_count>
+      \(
+        (?:
+          ## opt penalties
+            (?<pen>
+              (?:  (?<pen_value> \d{1,2}) [ ]? )?
+                 (?:pens|pen\.?|p)
+           )
+            |
+          ## opt own goals (og)
+            (?<og>
+             (?: (?<og_value> \d{1,2}) [ ]? )?
+                (?:ogs?|o\.g\.|o)
+            )
+            |
+          ## opt fallback - classic count/number
+          (?:  (?<value> [1-9])
+                ## check for option penalties
+                (?<pen>
+                     [,/] [ ]*
+                     (?: (?<pen_value> \d{1,2}) [ ]? )?
+                     (?:pens|pen\.?|p)
+                )?
+           )
+         )
+      \)
+)}ix
+
+
+
+
+##
+## note - inline \b check in MINUTE_RE excludes
+##      85pen  or 90+4pen or 38p
+##        (possible and NOT excluded in GOAL_MINUTE_RE  !!!)
+##
+##  minute with optional stoppage (offset)
+
+MINUTE_RE = %r{
+     (?<minute>
+               \b
+             (?<value>\d{1,3})      ## constrain numbers to 0 to 999!!!
+                \b
+                '?    ## optional minute marker
+
+                (?: \+ (?<value2>\d{1,2})
+                       \b
+                      '?    ## optional minute marker
+                 )?
+      )
+}ix
+
+
+
+##
+##  keep separate? or add simply inside GOAL_MINUTE_RE - why? why not?
+##   fix-fix-fix - move into GOAL_MINUTE_RE !!!
+
+GOAL_MINUTE_NA_RE = %r{
+     (?<goal_minute_na>
+
+       # positive lookbehind
+       (?<=[ ,;])
+
+       (?<value> \?{1,2})
+            '?    ## optional minute marker
+     ## note - add goal minute qualifiers here inline!!!
+        (?:
+            (?: [ ]? (?<og>   (?: \((?:og|o\.g\.|o)\))   ## allow (og)
+                                   |
+                              (?: (?:og|o\.g\.|o))      ## allow plain og
+                      )
+            )
+            |
+            (?: [ ]? (?<pen>  (?: \((?:pen\.?|p)\))   ## allow ()
+                                   |
+                              (?: (?:pen\.?|p))
+                      )
+            )
+            |
+            ## add experimental header qualifier
+            (?: [ ]? (?<hdr> \( (?:hdr\.?|h ) \) | (?: hdr\.?|h ) ))
+            |
+            ## add experimental free kick qualifier
+            (?: [ ]? (?<fk> \( (?:fk\.?|f ) \) | (?: fk\.?|f) ))
+        )?
+
+     ## note - check positive lookahead
+     (?=[ ,;)]|$)
+   )
+}ix
+
+
+##   goal types
+# (pen.) or (pen) or (p.) or (p)
+## (o.g.) or (og)
+##   todo/check - keep case-insensitive
+##                   or allow OG or P or PEN or
+##                   only lower case - why? why not?
+##
+##  add (gg) for golden goal - why? why not?
+##  add (sg) for silver goal - why? why not??
+
+GOAL_MINUTE_RE = %r{
+     (?<goal_minute>
+               \b
+             (?<value>\d{1,3})      ## constrain numbers to 0 to 999!!!
+                '?    ## optional minute marker
+
+                 (?: \+ (?<value2>\d{1,2})
+                      '?    ## optional minute marker
+                 )?
+
+        ## note - add goal minute qualifiers here inline!!!
+        (?:
+            (?: [ ]? (?<og>   (?: \((?:og|o\.g\.|o)\))   ## allow (og)
+                                   |
+                              (?: (?:og|o\.g\.|o))      ## allow plain og
+                      )
+            )
+            |
+            (?: [ ]? (?<pen>  (?: \((?:pen\.?|p)\))   ## allow ()
+                                   |
+                              (?: (?:pen\.?|p))
+                      )
+            )
+            |
+            ## add experimental header qualifier
+            (?: [ ]? (?<hdr> \( (?:hdr\.?|h ) \) | (?: hdr\.?|h ) ))
+            |
+            ## add experimental free kick qualifier
+            (?: [ ]? (?<fk> \( (?:fk\.?|f ) \) | (?: fk\.?|f) ))
+        )?
+
+        ##  add experimental seconds
+        ##    e.g. (95 secs) or (95sec) etc.
+        (?: [ ]*  \(
+                      (?<secs>\d{1,3})
+                         [ ]?secs?
+                   \)
+        )?
+     )
+
+     ## note - check positive lookahead
+     (?=[ ,;)]|$)
+}ix
+
+
+
+
+
+###
+## more regex  for goal alt
+
+
+GOAL_TYPE_RE = %r{
+     (?<goal_type>
+               \(
+                 (?:
+                      (?<og>  og|o\.g\.|o )
+                         |
+                      (?<pen> pen\.?|p )
+                         |
+                     ## add experimental header qualifier
+                      (?<hdr>  hdr\.?|h )
+                         |
+                     ## add experimental free kick qualifier
+                       (?<fk>  fk\.?|f )
+                  )
+                \)
+)}xi
+
+
+
+
+end  # class Lexer
+end # module Fbtxt

data/lib/fbtxt/lexer/token-group.rb

@@ -0,0 +1,29 @@
+module Fbtxt
+class Lexer
+
+
+###
+#   check for start of group def line e.g.
+#       Group A  | ...
+#       Group 1  : ....
+#       Group A2 | ....
+##  note - use \A (instead of ^) - \A strictly matches the start of the string.
+
+START_WITH_GROUP_DEF_LINE_RE =  %r{
+                     \A
+                     [ ]*  ## ignore leading spaces (if any)
+                     (?<group_def>
+                         Group
+                          [ ]
+                          [a-z0-9]+   ## todo/check - allow dot (.) too e.g. 1.A etc.- why? why not?
+                     )
+                     ###   positive lookahead MUST be : OR |
+                     (?= [ ]*
+                         [:|]
+                         [ ])  ## note: requires space for now after [:|] - keep - why? why not?
+                  }ix
+
+
+
+end  #   class Lexer
+end  # module Fbtxt

data/lib/fbtxt/lexer/token-note.rb

@@ -0,0 +1,40 @@
+module Fbtxt
+class Lexer
+
+
+### fix - use (?<text>) - text capture for inner text!!
+##         use (?<note> for complete match as a convention!! )
+NOTE_RE = %r{
+    \[
+      (?<note>
+         [^\[\]\#]*?    ## note - non-greedy/lazy operator
+                        ##    exclude comments inside note block - why? why not?
+      )
+    \]
+    }xi
+
+
+####
+##  fix - change NOTE_RE to MATCH_NOTE_RE !!!!
+##         and change NOTA_BENE_RE to NOTE_RE !!!
+
+
+
+## check for "literal"  (multi-line) note blocks
+##   eg.  nb:  or note:
+##   space required after double colon - why? why not?
+##
+##  note - use \A (instead of ^) - \A strictly matches the start of the string.
+NOTA_BENE_RE =  %r{   \A
+                           [ ]*  ## ignore leading spaces (if any)
+                        (?: nb | note) [ ]* : [ ]+
+                         (?<nota_bene>
+                              .+?  ## use non-greedy
+                          )
+                           [ ]*  ## ignore trailing spaces (if any)
+                          \z
+                       }xi
+
+
+end  #  class Lexer
+end  # module Fbtxt

data/lib/fbtxt/lexer/token-prop.rb

@@ -0,0 +1,334 @@
+###
+##  team prop mode e.g.
+##
+##
+##    Fri Jun 14 21:00  @ München Fußball Arena, München
+##     Germany  v  Scotland   5-1 (3-0)
+##  (Wirtz 10' Musiala 19' Havertz 45+1' (pen.) Füllkrug 68' Can 90+3'; Rüdiger 87' (o.g.))
+##
+## Germany:    Neuer - Kimmich, Rüdiger, Tah [Y], Mittelstädt - Andrich [Y] (Groß 46'),
+##       Kroos (Can 80') - Musiala (Müller 74'), Gündogan, Wirtz (Sane 63') -
+##       Havertz (Füllkrug 63')
+## Scotland:   Gunn - Porteous [R 44'], Hendry, Tierney (McKenna 78') - Ralston [Y],
+##       McTominay, McGregor (Gilmour 67'), Robertson - Christie (Shankland 82'),
+##       Adams (Hanley 46'), McGinn (McLean 67')
+
+
+module Fbtxt
+class Lexer
+
+
+##############
+#  add support for props/ attributes e.g.
+#
+#    Germany:    Neuer - Kimmich, Rüdiger, Tah [Y], Mittelstädt - Andrich [Y] (46' Groß),
+#      Kroos (80' Can) - Musiala (74' Müller), Gündogan,
+#      Wirtz (63' Sane) - Havertz (63' Füllkrug)
+#    Scotland:   Gunn - Porteous [R 44'], Hendry, Tierney (78' McKenna) - Ralston [Y],
+#      McTominay, McGregor (67' Gilmour), Robertson - Christie (82' Shankland),
+#      Adams (46' Hanley), McGinn (67' McLean)
+#
+## note:  colon (:) MUST be followed by one (or more) spaces
+##      make sure mon feb 12 18:10 will not match
+##        allow 1. FC Köln etc.
+##               Mainz 05:
+##           limit to 30 chars max
+##          only allow  chars incl. intl but (NOT ()[]/;)
+##
+##
+##  note - use special \G - Matches first matching position !!!!
+##     check for \G like backreference of regex tokens/parts if possible/available in ruby?
+
+
+
+  ## (i) starting w/ letters
+  ##                   note - incl./allows digits (0-9)
+  ##          e.g. a1, a2000, etc.
+  ##
+  ## note - added back optional trailing dot (.) for abbrev. word  !!!
+  PROP_KEY_WORD_ = %r{
+                           \p{L}
+                             [\p{L}\d]*
+                             \.?
+                    }ix
+
+  ## note - incl. optional dot or numsign e.g. 1. or 1°
+  PROP_KEY_NUM_ = %r{
+                              \d+
+                              [.°]?
+                   }ix
+
+  ## e.g. 1A, 1FC etc.
+  ##  note - no trailing dot (.) for now - check if any cases exist in real world
+  PROP_KEY_NUMALPHA_ = %r{
+                              \d+
+                              \p{L}
+                               [\p{L}\d]*
+                     }ix
+
+
+
+
+
+  START_WITH_PROP_KEY_RE = %r{
+                   \A         ## note - MUST start line; leading spaces optional (eat-up)
+                 (?<prop_key>
+                     [ ]*     ##  optional leading spaces
+                   (?<key>
+                       (?:
+                           ## (i) starting w/ letters
+                             #{PROP_KEY_WORD_}
+
+                           ## (ii) starting w/ number
+                           ##  e.g. 1fc, 1a,
+                           | #{PROP_KEY_NUMALPHA_}
+                           ##      followed by optional dot) and
+                           ##                  optional space
+                           ##      MUST be follow by letter (a to z)!!!!
+                           ##   eg. 1[ fc], 1.[ fc], 1.[fc],  etc.
+                           | #{PROP_KEY_NUM_}   (?= [ ]? \p{L})
+                       )
+                       (?:
+                           ## connectors  - note - no dot (.), must match with abbrev word or num!!
+                            (?: ## (i)   single space or WITHOUT surrounding spaces!! - slash (/), dash (-)
+                                ##     e.g. do NOT match   one - two     or one / two
+                                ##                        only one-two   or  one/two
+
+                                  [ /-]
+
+                                ## (ii)     surrounded by leading or trailing optional space
+                                ##            c & a, etc.
+                                ##            d'ivoire, d' ivoire
+                                ##            borusia 'gladbach etc.
+                                ##              exclude space ' space - why? why not? (or ignore for now)
+                                ##
+                                ##    check for quotes  ('') - not realy supported here
+                                ##              e.g. leading or trailing ' will NOT match
+
+                                 |  [ ]? & [ ]?
+                                 |  [ ]? '
+                                 |  ' [ ]?
+
+                                #### (iii)
+                                ##   note - special "hack"  to connect WITHOUT space
+                                ##     for   Union 1.FC  and SKN St.Pölten or St.Pölten
+                                ##       connects      1.FC      => NUM+WORD
+                                ##                     1°Mayo    => NUM+WORD
+                                ##                     St.Pölten => ABBREV+WORD
+                                ##
+                                ## note - match WITHOUT (space) connector
+                                ##                  1.FC  (Union 1.FC Stein)
+                                ##               [WORD: "Union"], [NUM: "1."], [WORD: "FC"]
+                                ##                  St.Pölten (SKN St.Pölten)
+                                ##                [WORD: "SKN"], [ABBREV: "St."], [WORD: "Pölten"]
+                                |   (?<=  [.°] )
+                                    (?=  \p{L})
+                            )
+                             (?:
+                                   #{PROP_KEY_NUMALPHA_}
+                                |  #{PROP_KEY_NUM_}
+                                |  #{PROP_KEY_WORD_}
+                               )
+                       )*
+                      )       ## close <key> capture
+                    [ ]*?     ## slurp trailing spaces
+                     :
+
+                ## positive lookahead (must be followed by space!!)
+                ##     or allow end-of-line too
+                    (?= [ ]+|$)
+                   )  ## close <prop_key> capture
+                 }ix
+
+
+
+#####
+#  (tactics) formation   - possible with two numbers (e.g. 5-5 ?? or 8-2)
+##   e.g   (4-3-3)    or (433)
+##         (4-1-4-1)  or (4141)
+##         etc.
+
+FORMATION_RE = %r{
+                (?<formation>
+                    \(
+                       (?<value>
+                            (?: [1-9] (?: -[1-9]){1,5} )
+                          | (?: [1-9]{2,6})
+                        )
+                    \)
+                )
+              }x
+
+
+
+
+################
+##     todo/check - use token for card short cuts?
+##                if m[:name] == 'Y'
+##                 [:YELLOW_CARD, m[:name]]
+##               elsif m[:name] == 'R'
+##                 [:RED_CARD, m[:name]]
+##           -  [Y], [R], [Y/R]  Yellow-Red Card
+##    check if minutes possible inside [Y 46']
+##     add [c] for captain too
+
+
+##  [c] or [C] for marking player as captain
+##   support [y ] too - or require Y - why? why not?
+  INLINE_CAPTAIN = %r{ (?<inline_captain>
+                           \[ [cC] \]
+                       )}x
+
+  INLINE_YELLOW  = %r{ (?<inline_yellow>
+                          \[ [yY]
+                              ## optional minute
+                              (?: [ ]+
+                                (?<minute> \d{1,3})
+                                   '?
+                                (?:
+                                   \+
+                                   (?<offset>\d{1,2})
+                                    '?
+                                )?
+                              )?
+                          \]
+                     )}x
+
+  INLINE_RED  = %r{ (?<inline_red>
+                          \[ [rR]
+                              ## optional minute
+                              (?: [ ]+
+                                (?<minute> \d{1,3})
+                                   '?
+                                (?:
+                                   \+
+                                   (?<offset>\d{1,2})
+                                    '?
+                                )?
+                              )?
+                          \]
+                     )}x
+
+  ##
+  ##  note make slash (/) optionals
+  ##    e.g. y/r,yr or Y/R,YR
+
+  INLINE_YELLOW_RED  = %r{ (?<inline_yellow_red>
+                          \[  (?: y/?r | Y/?R)
+                              ## optional minute
+                              (?: [ ]+
+                                (?<minute> \d{1,3})
+                                   '?
+                                (?:
+                                   \+
+                                   (?<offset>\d{1,2})
+                                    '?
+                                )?
+                              )?
+                          \]
+                     )}x
+
+
+
+
+### simple prop key for inline use e.g.
+###    Coach:  or Trainer:  or ...  add more here later
+
+  PROP_KEY_INLINE_RE = %r{
+                    \b
+                 (?<prop_key>    ## note: use prop_key (NOT prop_key_inline or such)
+                   (?<key>
+                       \p{L}+
+                   )
+                    ## note - NO spaces allowed for key for now!!!
+                     :
+                    ## possitive lookahead (must be followed by space!!)
+                    (?=[ ]+)
+                   )
+                 }ix
+
+
+
+## note allow underscore inline e.g.
+##  5_000
+## discuss/check - allow space inline (e.g. 5 000) - why? why not?
+
+PROP_NUM_RE = %r{
+             \b
+              (?<num>
+                  (?<value> [0-9]+
+                             (?: _ [0-9]+)*
+                  )
+              )
+             \b
+            }x
+
+
+### todo/fix - allow more chars in enclosed name  - why? why not?
+##                     e.g.  (') - Cote D'Ivore etc.
+##  change to PAREN_NAME or PARENTHESIS or such - why? why not?
+ENCLOSED_NAME_RE = %r{
+                 (?<enclosed_name>
+                    \(
+                   (?<name>
+                       \p{L}+
+                       (?:
+                          [ ]
+                            \p{L}+
+                       )*
+                   )
+                     \)
+                 )
+         }ix
+
+
+
+
+###
+##  e.g.   -; ...   or
+##         None - ... or
+CARDS_NONE_LEFT_RE = %r{ (?<cards_none_left>
+                         ## (i) w/ semicolon (;)
+                            (?:  (?: - | \bnone)
+                                  [ ]* ;
+                            )
+                         ## (ii) w/ dash (-)
+                          | (?:   (?: ∅ | \bnone)
+                                  [ ]+ -        ##    space REQUIRED
+                                  (?=[ ]|\z)    ## positive lookahead - speace required
+                            )
+                    )}ix
+
+## e.g   ... ;-
+##    or ... ; None
+CARDS_NONE_RIGHT_RE = %r{ (?<cards_none_right>
+                        ## (i) w/ semicolon (;)
+                           (?:   ; [ ]*
+                               (?:   - | none\b)
+                           )
+                        ## (ii) w/ dash (-)
+                          | (?:
+                               (?<=[ ])      ## positive lookbehind - space required
+                                - [ ]+       ##    space REQUIRED
+                               (?:  ∅ | none\b)
+                            )
+                      )}ix
+
+
+CARDS_SEP_ALT_RE = %r{(?<cards_sep_alt>
+              (?<=[ ])      ## positive lookbehind - space required
+                -
+              (?=[ ]|\z)    ## positive lookahead - speace required
+             )}x
+
+PENALTIES_SEP_ALT_RE = %r{(?<penalties_sep_alt>
+              (?<=[ ])      ## positive lookbehind - space required
+                -
+              (?=[ ]|\z)    ## positive lookahead - speace required
+             )}x
+
+
+
+
+end  # class Lexer
+end  # module Fbtxt