[comment {-*- tcl -*- doctools manpage}]
[manpage_begin grammar::me::tcl n 0.2]
[keywords grammar]
[keywords parsing]
[keywords {virtual machine}]
[copyright {2005 Andreas Kupries <andreas_kupries@users.sourceforge.net>}]
[moddesc   {Grammar operations and usage}]
[titledesc {Virtual machine implementation I for parsing token streams}]
[category  {Grammars and finite automata}]
[require Tcl "8.5 9"]
[require grammar::me::tcl [opt 0.2]]
[description]
[para]

This package provides an implementation of the ME virtual machine.
Please go and read the document [syscmd grammar::me_intro] first if
you do not know what a ME virtual machine is.

[para]

This implementation is tied very strongly to Tcl. All the stacks in
the machine state are handled through the Tcl stack, all control flow
is handled by Tcl commands, and the remaining machine instructions are
directly mapped to Tcl commands. Especially the matching of
nonterminal symbols is handled by Tcl procedures as well, essentially
extending the machine implementation with custom instructions.

[para]

Further on the implementation handles only a single machine which is
uninteruptible during execution and hardwired for pull operation. I.e.
it explicitly requests each new token through a callback, pulling them
into its state.

[para]

A related package is [package grammar::peg::interp] which provides a
generic interpreter / parser for parsing expression grammars (PEGs),
implemented on top of this implementation of the ME virtual machine.

[section {API}]

The commands documented in this section do not implement any of the
instructions of the ME virtual machine. They provide the facilities
for the initialization of the machine and the retrieval of important
information.

[list_begin definitions]

[call [cmd ::grammar::me::tcl] [method cmd] [arg ...]]

This is an ensemble command providing access to the commands listed in
this section. See the methods themselves for detailed specifications.

[call [cmd ::grammar::me::tcl] [method init] [arg nextcmd] [opt [arg tokmap]]]

This command (re)initializes the machine. It returns the empty
string. This command has to be invoked before any other command of
this package.

[para]

The command prefix [arg nextcmd] represents the input stream of
characters and is invoked by the machine whenever the a new character
from the stream is required. The instruction for handling this is
[term ict_advance].

The callback has to return either the empty list, or a list of 4
elements containing the token, its lexeme attribute, and its location
as line number and column index, in this order.

The empty list is the signal that the end of the input stream has been
reached. The lexeme attribute is stored in the terminal cache, but
otherwise not used by the machine.

[para]

The optional dictionary [arg tokmap] maps from tokens to integer
numbers. If present the numbers impose an order on the tokens, which
is subsequently used by [term ict_match_tokrange] to determine if a
token is in the specified range or not. If no token map is specified
the lexicographic order of th token names will be used instead. This
choice is especially asensible when using characters as tokens.

[call [cmd ::grammar::me::tcl] [method lc] [arg location]]

This command converts the location of a token given as offset in the
input stream into the associated line number and column index. The
result of the command is a 2-element list containing the two values,
in the order mentioned in the previous sentence.

This allows higher levels to convert the location information found in
the error status and the generated AST into more human readable data.

[para]

[emph Note] that the command is not able to convert locations which
have not been reached by the machine yet. In other words, if the
machine has read 7 tokens the command is able to convert the offsets
[const 0] to [const 6], but nothing beyond that. This also shows that
it is not possible to convert offsets which refer to locations before
the beginning of the stream.

[para]

After a call of [method init] the state used for the conversion is
cleared, making further conversions impossible until the machine has
read tokens again.

[call [cmd ::grammar::me::tcl] [method tok] [arg from] [opt [arg to]]]

This command returns a Tcl list containing the part of the input
stream between the locations [arg from] and [arg to] (both
inclusive). If [arg to] is not specified it will default to the value
of [arg from].

[para]

Each element of the returned list is a list of four elements, the
token, its associated lexeme, line number, and column index, in this
order.

In other words, each element has the same structure as the result of
the [arg nextcmd] callback given to [cmd ::grammar::me::tcl::init]

[para]

This command places the same restrictions on its location arguments as
[cmd ::grammar::me::tcl::lc].

[call [cmd ::grammar::me::tcl] [method tokens]]

This command returns the number of tokens currently known to the ME
virtual machine.

[call [cmd ::grammar::me::tcl] [method sv]]

This command returns the current semantic value [term SV] stored in
the machine. This is an abstract syntax tree as specified in the
document [syscmd grammar::me_ast], section [sectref-external {AST VALUES}].

[call [cmd ::grammar::me::tcl] [method ast]]

This method returns the abstract syntax tree currently at the top of
the AST stack of the ME virtual machine. This is an abstract syntax
tree as specified in the document [syscmd grammar::me_ast], section
[sectref-external {AST VALUES}].

[call [cmd ::grammar::me::tcl] [method astall]]

This method returns the whole stack of abstract syntax trees currently
known to the ME virtual machine. Each element of the returned list is
an abstract syntax tree as specified in the document

[syscmd grammar::me_ast], section [sectref-external {AST VALUES}].
The top of the stack resides at the end of the list.

[call [cmd ::grammar::me::tcl] [method ctok]]

This method returns the current token considered by the ME virtual
machine.

[call [cmd ::grammar::me::tcl] [method nc]]

This method returns the contents of the nonterminal cache as a
dictionary mapping from "[var symbol],[var location]" to match
information.

[call [cmd ::grammar::me::tcl] [method next]]

This method returns the next token callback as specified during
initialization of the ME virtual machine.

[call [cmd ::grammar::me::tcl] [method ord]]

This method returns a dictionary containing the [arg tokmap] specified
during initialization of the ME virtual machine.

[var [cmd ::grammar::me::tcl::ok]]

This variable contains the current match status [term OK]. It is
provided as variable instead of a command because that makes access to
this information faster, and the speed of access is considered very
important here as this information is used constantly to determine the
control flow.

[list_end]
[para]

[section {MACHINE STATE}]

Please go and read the document [syscmd grammar::me_vm] first for a
specification of the basic ME virtual machine and its state.

[para]

This implementation manages the state described in that document,
except for the stacks minus the AST stack. In other words, location
stack, error stack, return stack, and ast marker stack are implicitly
managed through standard Tcl scoping, i.e. Tcl variables in
procedures, outside of this implementation.

[section {MACHINE INSTRUCTIONS}]

Please go and read the document [syscmd grammar::me_vm] first for a
specification of the basic ME virtual machine and its instruction set.

[para]

This implementation maps all instructions to Tcl commands in the
namespace "::grammar::me::tcl", except for the stack related commands,
nonterminal symbols and control flow.

Here we simply list the commands and explain the differences to the
specified instructions, if there are any. For their semantics see the
aforementioned specification. The machine commands are [emph not]
reachable through the ensemble command [cmd ::grammar::me::tcl].

[para]
[list_begin definitions]

[call [cmd ::grammar::me::tcl::ict_advance] [arg message]]
No changes.

[call [cmd ::grammar::me::tcl::ict_match_token] [arg tok] [arg message]]
No changes.

[call [cmd ::grammar::me::tcl::ict_match_tokrange] [arg tokbegin] [arg tokend] [arg message]]

If, and only if a token map was specified during initialization then
the arguments are the numeric representations of the smallest and
largest tokens in the range. Otherwise they are the relevant tokens
themselves and lexicographic comparison is used.

[call [cmd ::grammar::me::tcl::ict_match_tokclass] [arg code] [arg message]]
No changes.

[call [cmd ::grammar::me::tcl::inc_restore] [arg nt]]

Instead of taking a branchlabel the command returns a boolean value.
The result will be [const true] if and only if cached information was
found. The caller has to perform the appropriate branching.

[call [cmd ::grammar::me::tcl::inc_save] [arg nt] [arg startlocation]]

The command takes the start location as additional argument, as it is
managed on the Tcl stack, and not in the machine state.

[def "[cmd icf_ntcall] [arg branchlabel]"]
[def [cmd icf_ntreturn]]

These two instructions are not mapped to commands. They are control
flow instructions and handled in Tcl.

[call [cmd ::grammar::me::tcl::iok_ok]]
No changes.

[call [cmd ::grammar::me::tcl::iok_fail]]
No changes.

[call [cmd ::grammar::me::tcl::iok_negate]]
No changes.

[def "[cmd icf_jalways] [arg branchlabel]"]
[def "[cmd icf_jok] [arg branchlabel]"]
[def "[cmd icf_jfail] [arg branchlabel]"]
[def [cmd icf_halt]]

These four instructions are not mapped to commands. They are control
flow instructions and handled in Tcl.

[call [cmd ::grammar::me::tcl::icl_get]]

This command returns the current location [term CL] in the input.
It replaces [term icl_push].

[call [cmd ::grammar::me::tcl::icl_rewind] [arg oldlocation]]

The command takes the location as argument as it comes from the
Tcl stack, not the machine state.

[def [cmd icl_pop]]

Not mapped, the stacks are not managed by the package.

[call [cmd ::grammar::me::tcl::ier_get]]

This command returns the current error state [term ER].
It replaces [term ier_push].

[call [cmd ::grammar::me::tcl::ier_clear]]
No changes.

[call [cmd ::grammar::me::tcl::ier_nonterminal] [arg message] [arg location]]

The command takes the location as argument as it comes from the
Tcl stack, not the machine state.

[call [cmd ::grammar::me::tcl::ier_merge] [arg olderror]]

The command takes the second error state to merge as argument as it
comes from the Tcl stack, not the machine state.

[call [cmd ::grammar::me::tcl::isv_clear]]
No changes.

[call [cmd ::grammar::me::tcl::isv_terminal]]
No changes.

[call [cmd ::grammar::me::tcl::isv_nonterminal_leaf] [arg nt] [arg startlocation]]

The command takes the start location as argument as it comes from the
Tcl stack, not the machine state.

[call [cmd ::grammar::me::tcl::isv_nonterminal_range] [arg nt] [arg startlocation]]

The command takes the start location as argument as it comes from the
Tcl stack, not the machine state.

[call [cmd ::grammar::me::tcl::isv_nonterminal_reduce] [arg nt] [arg startlocation] [opt [arg marker]]]

The command takes start location and marker as argument as it comes
from the Tcl stack, not the machine state.

[call [cmd ::grammar::me::tcl::ias_push]]
No changes.

[call [cmd ::grammar::me::tcl::ias_mark]]

This command returns a marker for the current state of the AST stack
[term AS]. The marker stack is not managed by the machine.

[call [cmd ::grammar::me::tcl::ias_pop2mark] [arg marker]]

The command takes the marker as argument as it comes from the
Tcl stack, not the machine state. It replaces [term ias_mpop].

[list_end]
[para]

[vset CATEGORY grammar_me]
[include ../common-text/feedback.inc]
[manpage_end]