lark: Question (possibly bug?)

Hello Erez,

I’m having trouble coming up with the correct rule when parsing the sudo grammar. Here’s my grammar:

      ?sudo_item : (alias | user_spec)*

      alias : "User_Alias"    user_alias  (":" user_alias)*
              | "Runas_Alias" runas_alias (":" runas_alias)*
              | "Host_Alias"  host_alias  (":" host_alias)*
              | "Cmnd_Alias"  cmnd_alias  (":" cmnd_alias)*

      user_alias  : ALIAS_NAME "=" user_list
      host_alias  : ALIAS_NAME "=" host_list
      runas_alias : ALIAS_NAME "=" runas_list
      cmnd_alias  : ALIAS_NAME "=" cmnd_list

      user_spec      : user_list host_list "=" cmnd_spec_list (":" host_list "=" cmnd_spec_list)*
      cmnd_spec_list : cmnd_spec ("," cmnd_spec)*
      cmnd_spec      : runas_spec? tag_spec* command
      runas_spec     : "(" runas_list? (":" runas_list)? ")"
      command        : "!"* command_name | "!"* cmnd_alias
      command_name   : COMMAND
      file_name      : /[-_.a-z0-9A-Z\/]\/]+/

      host_list  : host      ("," host)*
      user_list  : user      ("," user)*
      runas_list : user      ("," user)*
      cmnd_list  : command   ("," command)*

      host : HOST_NAME

      user :    "!"*       USER_NAME
              | "!"* "%"   GROUP_NAME
              | "!"* "#"   UID
              | "!"* "%#"  GID
              | "!"* "+"   NETGROUP_NAME
              | "!"* "%:"  NONUNIX_GROUP_NAME
              | "!"* "%:#" NONUNIX_GID
              | "!"* user_alias

      tag_spec : (tag_nopwd
                  | tag_pwd
                  | tag_noexec
                  | tag_exec
                  | tag_setenv
                  | tag_nosetenv
                  | tag_log_output
                  | tag_nolog_output) ":"

      tag_pwd          : "PASSWD"
      tag_nopwd        : "NOPASSWD"

      tag_exec         : "EXEC"
      tag_noexec       : "NOEXEC"
      tag_setenv       : "SETENV"
      tag_nosetenv     : "NOSETENV"
      tag_log_output   : "LOG_OUTPUT"
      tag_nolog_output : "NOLOG_OUTPUT"

      UID                : /[0-9]+/
      GID                : /[0-9]+/
      NONUNIX_GID        : /[0-9]+/
      USER_NAME          : /[-_.a-z0-9A-Z]+/
      GROUP_NAME         : CNAME
      NETGROUP_NAME      : CNAME
      NONUNIX_GROUP_NAME : CNAME
      ALIAS_NAME         : CNAME
      HOST_NAME          : /[-_.a-z0-9A-Z\[\]*]+/
      COMMAND            : /[^,:\n]+/

      %import common.CNAME
      %import common.WS
      %ignore /[\\\\]/
      %ignore WS

A sudo rule can look like this:

DBA ALL = (oracle) ALL, !SU

or like this:

DBA ALL = (oracle) ALL, !SU : ALL = (postgres) ALL, !SU

The grammar handles the first case fine. It has trouble parsing the second variant. It boils down to the COMMAND token, which is anything following the runas bit ((postgres), (oracle), etc) and should include anything but [:,\n]. That regex doesn’t seem to work.

Also, sudo lines may have the \ line continuation at the end. Is %ignore /[\\\\]/ the correct way to handle that situation? It does seem to operate correctly.

Last question, when i specify parser='lalr', I get this:

MY INPUT: DBA ALL = (oracle) ALL ERROR:

lark.common.UnexpectedToken: Unexpected token Token(COMMAND, ' ALL = (oracle) ALL') at line 1, column 3.

Does that mean my grammar is not LALR compatible?