msyds/gf-core
1
0
Fork 0
forked from GitHub/gf-core
Code Pull Requests Activity

documented new GFCC

Browse Source
This commit is contained in:
aarne
2007-10-05 16:18:23 +00:00
parent 2905d5552c
commit 51ac00a987
4 changed files with 1164 additions and 489 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/GF/GFCC/GFCC.cf
View File

@@ -29,11 +29,11 @@ Lin. LinDef ::= CId "=" Term ;
DTyp. Type ::= "[" [Hypo] "]" CId [Exp] ; -- dependent type DTyp. Type ::= "[" [Hypo] "]" CId [Exp] ; -- dependent type
DTr. Exp ::= "[" "(" [CId] ")" Atom [Exp] "]" ; -- term with bindings DTr. Exp ::= "[" "(" [CId] ")" Atom [Exp] "]" ; -- term with bindings
AC. Atom ::= CId ; AC. Atom ::= CId ;
AS. Atom ::= String ; AS. Atom ::= String ;
AI. Atom ::= Integer ; AI. Atom ::= Integer ;
AF. Atom ::= Double ; AF. Atom ::= Double ;
AM. Atom ::= "?" Integer ; AM. Atom ::= "?" Integer ;
R. Term ::= "[" [Term] "]" ; -- record/table R. Term ::= "[" [Term] "]" ; -- record/table
P. Term ::= "(" Term "!" Term ")" ; -- projection/selection P. Term ::= "(" Term "!" Term ")" ; -- projection/selection

557
src/GF/GFCC/doc/gfcc.html
View File

@@ -7,41 +7,9 @@
<P ALIGN="center"><CENTER><H1>The GFCC Grammar Format</H1> <P ALIGN="center"><CENTER><H1>The GFCC Grammar Format</H1>
<FONT SIZE="4"> <FONT SIZE="4">
<I>Aarne Ranta</I><BR> <I>Aarne Ranta</I><BR>
October 19, 2006 October 5, 2007
</FONT></CENTER> </FONT></CENTER>
<P></P>
<HR NOSHADE SIZE=1>
<P></P>
<UL>
<LI><A HREF="#toc1">What is GFCC</A>
<LI><A HREF="#toc2">GFCC vs. GFC</A>
<LI><A HREF="#toc3">The syntax of GFCC files</A>
<UL>
<LI><A HREF="#toc4">Top level</A>
<LI><A HREF="#toc5">Abstract syntax</A>
<LI><A HREF="#toc6">Concrete syntax</A>
</UL>
<LI><A HREF="#toc7">The semantics of concrete syntax terms</A>
<UL>
<LI><A HREF="#toc8">Linearization and realization</A>
<LI><A HREF="#toc9">Term evaluation</A>
<LI><A HREF="#toc10">The special term constructors</A>
</UL>
<LI><A HREF="#toc11">Compiling to GFCC</A>
<UL>
<LI><A HREF="#toc12">Problems in GFCC compilation</A>
<LI><A HREF="#toc13">The representation of linearization types</A>
<LI><A HREF="#toc14">Running the compiler and the GFCC interpreter</A>
</UL>
<LI><A HREF="#toc15">The reference interpreter</A>
<LI><A HREF="#toc16">Interpreter in C++</A>
<LI><A HREF="#toc17">Some things to do</A>
</UL>
<P></P>
<HR NOSHADE SIZE=1>
<P></P>
<P> <P>
Author's address: Author's address:
<A HREF="http://www.cs.chalmers.se/~aarne"><CODE>http://www.cs.chalmers.se/~aarne</CODE></A> <A HREF="http://www.cs.chalmers.se/~aarne"><CODE>http://www.cs.chalmers.se/~aarne</CODE></A>
@@ -50,11 +18,11 @@ Author's address:
History: History:
</P> </P>
<UL> <UL>
<LI>5 Oct 2007: new, better structured GFCC with full expressive power
<LI>19 Oct: translation of lincats, new figures on C++ <LI>19 Oct: translation of lincats, new figures on C++
<LI>3 Oct 2006: first version <LI>3 Oct 2006: first version
</UL> </UL>
<A NAME="toc1"></A>
<H2>What is GFCC</H2> <H2>What is GFCC</H2>
<P> <P>
GFCC is a low-level format for GF grammars. Its aim is to contain the minimum GFCC is a low-level format for GF grammars. Its aim is to contain the minimum
@@ -68,18 +36,20 @@ advantages:
</UL> </UL>
<P> <P>
The idea is that all embedded GF applications are compiled to GFCC. Thus we also want to call GFCC the <B>portable grammar format</B>.
</P>
<P>
The idea is that all embedded GF applications use GFCC.
The GF system would be primarily used as a compiler and as a grammar The GF system would be primarily used as a compiler and as a grammar
development tool. development tool.
</P> </P>
<P> <P>
Since GFCC is implemented in BNFC, a parser of the format is readily Since GFCC is implemented in BNFC, a parser of the format is readily
available for C, C++, Haskell, Java, and OCaml. Also an XML available for C, C++, C#, Haskell, Java, and OCaml. Also an XML
representation is generated in BNFC. A representation can be generated in BNFC. A
<A HREF="../">reference implementation</A> <A HREF="../">reference implementation</A>
of linearization and some other functions has been written in Haskell. of linearization and some other functions has been written in Haskell.
</P> </P>
<A NAME="toc2"></A>
<H2>GFCC vs. GFC</H2> <H2>GFCC vs. GFC</H2>
<P> <P>
GFCC is aimed to replace GFC as the run-time grammar format. GFC was designed GFCC is aimed to replace GFC as the run-time grammar format. GFC was designed
@@ -92,7 +62,14 @@ run-time. In particular, the pattern matching syntax and semantics of GFC is
complex and therefore difficult to implement in new platforms. complex and therefore difficult to implement in new platforms.
</P> </P>
<P> <P>
The main differences of GFCC compared with GFC can be summarized as follows: Actually, GFC is planned to be omitted also as the target format of
separate compilation, where plain GF (type annotated and partially evaluated)
will be used instead. GFC provides only marginal advantages as a target format
compared with GF, and it is therefore just extra weight to carry around this
format.
</P>
<P>
The main differences of GFCC compared with GFC (and GF) can be summarized as follows:
</P> </P>
<UL> <UL>
<LI>there are no modules, and therefore no qualified names <LI>there are no modules, and therefore no qualified names
@@ -101,56 +78,56 @@ The main differences of GFCC compared with GFC can be summarized as follows:
<LI>records and tables are replaced by arrays <LI>records and tables are replaced by arrays
<LI>record labels and parameter values are replaced by integers <LI>record labels and parameter values are replaced by integers
<LI>record projection and table selection are replaced by array indexing <LI>record projection and table selection are replaced by array indexing
<LI>there is (so far) no support for dependent types or higher-order abstract <LI>even though the format does support dependent types and higher-order abstract
syntax (which would be easy to add, but make interpreters much more difficult syntax, there is no interpreted yet that does this
to write)
</UL> </UL>
<P> <P>
Here is an example of a GF grammar, consisting of three modules, Here is an example of a GF grammar, consisting of three modules,
as translated to GFCC. The representations are aligned, with the exceptions as translated to GFCC. The representations are aligned; thus they do not completely
due to the alphabetical sorting of GFCC grammars. reflect the order of judgements in GFCC files, which have different orders of
blocks of judgements, and alphabetical sorting.
</P> </P>
<PRE> <PRE>
grammar Ex(Eng,Swe); grammar Ex(Eng,Swe);
abstract Ex = { abstract { abstract Ex = { abstract {
cat cat cat
S ; NP ; VP ; S ; NP ; VP ; NP[]; S[]; VP[];
fun fun fun
Pred : NP -&gt; VP -&gt; S ; Pred : NP,VP -&gt; S = (Pred); Pred : NP -&gt; VP -&gt; S ; Pred=[(($ 0! 1),(($ 1! 0)!($ 0! 0)))];
She, They : NP ; She : -&gt; NP = (She); She, They : NP ; She=[0,"she"];
Sleep : VP ; Sleep : -&gt; VP = (Sleep); Sleep : VP ; They=[1,"they"];
They : -&gt; NP = (They); Sleep=[["sleeps","sleep"]];
} } ; } } ;
concrete Eng of Ex = { concrete Eng { concrete Eng of Ex = { concrete Eng {
lincat lincat lincat
S = {s : Str} ; S = {s : Str} ; S=[()];
NP = {s : Str ; n : Num} ; NP = {s : Str ; n : Num} ; NP=[1,()];
VP = {s : Num =&gt; Str} ; VP = {s : Num =&gt; Str} ; VP=[[(),()]];
param param
Num = Sg | Pl ; Num = Sg | Pl ;
lin lin lin
Pred np vp = { Pred = [(($0!1),(($1!0)!($0!0)))]; Pred np vp = { Pred=[(($ 0! 1),(($ 1! 0)!($ 0! 0)))];
s = np.s ++ vp.s ! np.n} ; s = np.s ++ vp.s ! np.n} ;
She = {s = "she" ; n = Sg} ; She = [0, "she"]; She = {s = "she" ; n = Sg} ; She=[0,"she"];
They = {s = "they" ; n = Pl} ; They = {s = "they" ; n = Pl} ; They = [1, "they"];
Sleep = {s = table { Sleep = [("sleep" + ["s",""])]; Sleep = {s = table { Sleep=[["sleeps","sleep"]];
Sg =&gt; "sleeps" ; Sg =&gt; "sleeps" ;
Pl =&gt; "sleep" They = [1, "they"]; Pl =&gt; "sleep"
} } ; }
} ; } ;
} } } ;
concrete Swe of Ex = { concrete Swe { concrete Swe of Ex = { concrete Swe {
lincat lincat lincat
S = {s : Str} ; S = {s : Str} ; S=[()];
NP = {s : Str} ; NP = {s : Str} ; NP=[()];
VP = {s : Str} ; VP = {s : Str} ; VP=[()];
param param
Num = Sg | Pl ; Num = Sg | Pl ;
lin lin lin
Pred np vp = { Pred = [(($0!0),($1!0))]; Pred np vp = { Pred = [(($0!0),($1!0))];
s = np.s ++ vp.s} ; s = np.s ++ vp.s} ;
She = {s = "hon"} ; She = ["hon"]; She = {s = "hon"} ; She = ["hon"];
@@ -159,9 +136,12 @@ due to the alphabetical sorting of GFCC grammars.
} } ; } } ;
</PRE> </PRE>
<P></P> <P></P>
<A NAME="toc3"></A>
<H2>The syntax of GFCC files</H2> <H2>The syntax of GFCC files</H2>
<A NAME="toc4"></A> <P>
The complete BNFC grammar, from which
the rules in this section are taken, is in the file
<A HREF="../DataGFCC.cf"><CODE>GF/GFCC/GFCC.cf</CODE></A>.
</P>
<H3>Top level</H3> <H3>Top level</H3>
<P> <P>
A grammar has a header telling the name of the abstract syntax A grammar has a header telling the name of the abstract syntax
@@ -170,25 +150,43 @@ the concrete languages. The abstract syntax and the concrete
syntaxes themselves follow. syntaxes themselves follow.
</P> </P>
<PRE> <PRE>
Grammar ::= Header ";" Abstract ";" [Concrete] ; Grm. Grammar ::=
Header ::= "grammar" CId "(" [CId] ")" ; "grammar" CId "(" [CId] ")" ";"
Abstract ::= "abstract" "{" [AbsDef] "}" ; Abstract ";"
Concrete ::= "concrete" CId "{" [CncDef] "}" ; [Concrete] ;
Abs. Abstract ::=
"abstract" "{"
"flags" [Flag]
"fun" [FunDef]
"cat" [CatDef]
"}" ;
Cnc. Concrete ::=
"concrete" CId "{"
"flags" [Flag]
"lin" [LinDef]
"oper" [LinDef]
"lincat" [LinDef]
"lindef" [LinDef]
"printname" [LinDef]
"}" ;
</PRE> </PRE>
<P> <P>
Abstract syntax judgements give typings and semantic definitions. This syntax organizes each module to a sequence of <B>fields</B>, such
Concrete syntax judgements give linearizations. as flags, linearizations, operations, linearization types, etc.
It is envisaged that particular applications can ignore some
of the fields, typically so that earlier fields are more
important than later ones.
</P> </P>
<PRE>
AbsDef ::= CId ":" Type "=" Exp ;
CncDef ::= CId "=" Term ;
</PRE>
<P> <P>
Also flags are possible, local to each "module" (i.e. abstract and concretes). The judgement forms have the following syntax.
</P> </P>
<PRE> <PRE>
AbsDef ::= "%" CId "=" String ; Flg. Flag ::= CId "=" String ;
CncDef ::= "%" CId "=" String ; Cat. CatDef ::= CId "[" [Hypo] "]" ;
Fun. FunDef ::= CId ":" Type "=" Exp ;
Lin. LinDef ::= CId "=" Term ;
</PRE> </PRE>
<P> <P>
For the run-time system, the reference implementation in Haskell For the run-time system, the reference implementation in Haskell
@@ -203,33 +201,84 @@ uses a structure that gives efficient look-up:
} }
data Abstr = Abstr { data Abstr = Abstr {
funs :: Map CId Type, -- find the type of a fun aflags :: Map CId String, -- value of a flag
cats :: Map CId [CId] -- find the funs giving a cat funs :: Map CId (Type,Exp), -- type and def of a fun
cats :: Map CId [Hypo], -- context of a cat
catfuns :: Map CId [CId] -- funs yielding a cat (redundant, for fast lookup)
} }
type Concr = Map CId Term data Concr = Concr {
flags :: Map CId String, -- value of a flag
lins :: Map CId Term, -- lin of a fun
opers :: Map CId Term, -- oper generated by subex elim
lincats :: Map CId Term, -- lin type of a cat
lindefs :: Map CId Term, -- lin default of a cat
printnames :: Map CId Term -- printname of a cat or a fun
}
</PRE>
<P>
These definitions are from <A HREF="../DataGFCC.hs"><CODE>GF/GFCC/DataGFCC.hs</CODE></A>.
</P>
<P>
Identifiers (<CODE>CId</CODE>) are like <CODE>Ident</CODE> in GF, except that
the compiler produces constants prefixed with <CODE>_</CODE> in
the common subterm elimination optimization.
</P>
<PRE>
token CId (('_' | letter) (letter | digit | '\'' | '_')*) ;
</PRE> </PRE>
<P></P> <P></P>
<A NAME="toc5"></A>
<H3>Abstract syntax</H3> <H3>Abstract syntax</H3>
<P> <P>
Types are first-order function types built from Types are first-order function types built from argument type
contexts and value types.
category symbols. Syntax trees (<CODE>Exp</CODE>) are category symbols. Syntax trees (<CODE>Exp</CODE>) are
rose trees with the head (<CODE>Atom</CODE>) either a function rose trees with nodes consisting of a head (<CODE>Atom</CODE>) and
constant, a metavariable, or a string, integer, or float bound variables (<CODE>CId</CODE>).
</P>
<PRE>
DTyp. Type ::= "[" [Hypo] "]" CId [Exp] ;
DTr. Exp ::= "[" "(" [CId] ")" Atom [Exp] "]" ;
Hyp. Hypo ::= CId ":" Type ;
</PRE>
<P>
The head Atom is either a function
constant, a bound variable, or a metavariable, or a string, integer, or float
literal. literal.
</P> </P>
<PRE> <PRE>
Type ::= [CId] "-&gt;" CId ; AC. Atom ::= CId ;
Exp ::= "(" Atom [Exp] ")" ; AS. Atom ::= String ;
Atom ::= CId ; -- function constant AI. Atom ::= Integer ;
Atom ::= "?" ; -- metavariable AF. Atom ::= Double ;
Atom ::= String ; -- string literal AM. Atom ::= "?" Integer ;
Atom ::= Integer ; -- integer literal
Atom ::= Double ; -- float literal
</PRE> </PRE>
<P></P> <P>
<A NAME="toc6"></A> The context-free types and trees of the "old GFCC" are special
cases, which can be defined as follows:
</P>
<PRE>
Typ. Type ::= [CId] "-&gt;" CId
Typ args val = DTyp [Hyp (CId "_") arg | arg &lt;- args] val
Tr. Exp ::= "(" CId [Exp] ")"
Tr fun exps = DTr [] fun exps
</PRE>
<P>
To store semantic (<CODE>def</CODE>) definitions by cases, the following expression
form is provided, but it is only meaningful in the last field of a function
declaration in an abstract syntax:
</P>
<PRE>
EEq. Exp ::= "{" [Equation] "}" ;
Equ. Equation ::= [Exp] "-&gt;" Exp ;
</PRE>
<P>
Notice that expressions are used to encode patterns. Primitive notions
(the default semantics in GF) are encoded as empty sets of equations
(<CODE>[]</CODE>). For a constructor (canonical form) of a category <CODE>C</CODE>, we
aim to use the encoding as the application <CODE>(_constr C)</CODE>.
</P>
<H3>Concrete syntax</H3> <H3>Concrete syntax</H3>
<P> <P>
Linearization terms (<CODE>Term</CODE>) are built as follows. Linearization terms (<CODE>Term</CODE>) are built as follows.
@@ -237,12 +286,12 @@ Constructor names are shown to make the later code
examples readable. examples readable.
</P> </P>
<PRE> <PRE>
R. Term ::= "[" [Term] "]" ; -- array R. Term ::= "[" [Term] "]" ; -- array (record/table)
P. Term ::= "(" Term "!" Term ")" ; -- access to indexed field P. Term ::= "(" Term "!" Term ")" ; -- access to field (projection/selection)
S. Term ::= "(" [Term] ")" ; -- sequence with ++ S. Term ::= "(" [Term] ")" ; -- concatenated sequence
K. Term ::= Tokn ; -- token K. Term ::= Tokn ; -- token
V. Term ::= "$" Integer ; -- argument V. Term ::= "$" Integer ; -- argument (subtree)
C. Term ::= Integer ; -- array index C. Term ::= Integer ; -- array index (label/parameter value)
FV. Term ::= "[|" [Term] "|]" ; -- free variation FV. Term ::= "[|" [Term] "|]" ; -- free variation
TM. Term ::= "?" ; -- linearization of metavariable TM. Term ::= "?" ; -- linearization of metavariable
</PRE> </PRE>
@@ -256,7 +305,7 @@ variant lists.
Var. Variant ::= [String] "/" [String] ; Var. Variant ::= [String] "/" [String] ;
</PRE> </PRE>
<P> <P>
Three special forms of terms are introduced by the compiler Two special forms of terms are introduced by the compiler
as optimizations. They can in principle be eliminated, but as optimizations. They can in principle be eliminated, but
their presence makes grammars much more compact. Their semantics their presence makes grammars much more compact. Their semantics
will be explained in a later section. will be explained in a later section.
@@ -264,20 +313,20 @@ will be explained in a later section.
<PRE> <PRE>
F. Term ::= CId ; -- global constant F. Term ::= CId ; -- global constant
W. Term ::= "(" String "+" Term ")" ; -- prefix + suffix table W. Term ::= "(" String "+" Term ")" ; -- prefix + suffix table
RP. Term ::= "(" Term "@" Term ")"; -- record parameter alias
</PRE> </PRE>
<P> <P>
Identifiers are like <CODE>Ident</CODE> in GF and GFC, except that There is also a deprecated form of "record parameter alias",
the compiler produces constants prefixed with <CODE>_</CODE> in
the common subterm elimination optimization.
</P> </P>
<PRE> <PRE>
token CId (('_' | letter) (letter | digit | '\'' | '_')*) ; RP. Term ::= "(" Term "@" Term ")"; -- DEPRECATED
</PRE> </PRE>
<P></P> <P>
<A NAME="toc7"></A> which will be removed when the migration to new GFCC is complete.
</P>
<H2>The semantics of concrete syntax terms</H2> <H2>The semantics of concrete syntax terms</H2>
<A NAME="toc8"></A> <P>
The code in this section is from <A HREF="../Linearize.hs"><CODE>GF/GFCC/Linearize.hs</CODE></A>.
</P>
<H3>Linearization and realization</H3> <H3>Linearization and realization</H3>
<P> <P>
The linearization algorithm is essentially the same as in The linearization algorithm is essentially the same as in
@@ -289,18 +338,21 @@ in which linearization is performed.
</P> </P>
<PRE> <PRE>
linExp :: GFCC -&gt; CId -&gt; Exp -&gt; Term linExp :: GFCC -&gt; CId -&gt; Exp -&gt; Term
linExp mcfg lang tree@(Tr at trees) = case at of linExp gfcc lang tree@(DTr _ at trees) = case at of
AC fun -&gt; comp (Prelude.map lin trees) $ look fun AC fun -&gt; comp (Prelude.map lin trees) $ look fun
AS s -&gt; R [kks (show s)] -- quoted AS s -&gt; R [kks (show s)] -- quoted
AI i -&gt; R [kks (show i)] AI i -&gt; R [kks (show i)]
AF d -&gt; R [kks (show d)] AF d -&gt; R [kks (show d)]
AM -&gt; TM AM -&gt; TM
where where
lin = linExp mcfg lang lin = linExp gfcc lang
comp = compute mcfg lang comp = compute gfcc lang
look = lookLin mcfg lang look = lookLin gfcc lang
</PRE> </PRE>
<P> <P>
TODO: bindings must be supported.
</P>
<P>
The result of linearization is usually a record, which is realized as The result of linearization is usually a record, which is realized as
a string using the following algorithm. a string using the following algorithm.
</P> </P>
@@ -316,12 +368,12 @@ a string using the following algorithm.
TM -&gt; "?" TM -&gt; "?"
</PRE> </PRE>
<P> <P>
Since the order of record fields is not necessarily Notice that realization always picks the first field of a record.
the same as in GF source, If a linearization type has more than one field, the first field
this realization does not work securely for does not necessarily contain the desired string.
categories whose lincats more than one field. Also notice that the order of record fields in GFCC is not necessarily
the same as in GF source.
</P> </P>
<A NAME="toc9"></A>
<H3>Term evaluation</H3> <H3>Term evaluation</H3>
<P> <P>
Evaluation follows call-by-value order, with two environments Evaluation follows call-by-value order, with two environments
@@ -339,10 +391,9 @@ deep patterns (such as Java and C++).
</P> </P>
<PRE> <PRE>
compute :: GFCC -&gt; CId -&gt; [Term] -&gt; Term -&gt; Term compute :: GFCC -&gt; CId -&gt; [Term] -&gt; Term -&gt; Term
compute mcfg lang args = comp where compute gfcc lang args = comp where
comp trm = case trm of comp trm = case trm of
P r p -&gt; proj (comp r) (comp p) P r p -&gt; proj (comp r) (comp p)
RP i t -&gt; RP (comp i) (comp t)
W s t -&gt; W s (comp t) W s t -&gt; W s (comp t)
R ts -&gt; R $ Prelude.map comp ts R ts -&gt; R $ Prelude.map comp ts
V i -&gt; idx args (fromInteger i) -- already computed V i -&gt; idx args (fromInteger i) -- already computed
@@ -351,7 +402,7 @@ deep patterns (such as Java and C++).
S ts -&gt; S $ Prelude.filter (/= S []) $ Prelude.map comp ts S ts -&gt; S $ Prelude.filter (/= S []) $ Prelude.map comp ts
_ -&gt; trm _ -&gt; trm
look = lookLin mcfg lang look = lookOper gfcc lang
idx xs i = xs !! i idx xs i = xs !! i
@@ -377,7 +428,6 @@ deep patterns (such as Java and C++).
_ -&gt; trace ("ERROR in grammar compiler: field from " ++ show t) t _ -&gt; trace ("ERROR in grammar compiler: field from " ++ show t) t
</PRE> </PRE>
<P></P> <P></P>
<A NAME="toc10"></A>
<H3>The special term constructors</H3> <H3>The special term constructors</H3>
<P> <P>
The three forms introduced by the compiler may a need special The three forms introduced by the compiler may a need special
@@ -391,13 +441,13 @@ Global constants
</PRE> </PRE>
<P> <P>
are shorthands for complex terms. They are produced by the are shorthands for complex terms. They are produced by the
compiler by (iterated) common subexpression elimination. compiler by (iterated) <B>common subexpression elimination</B>.
They are often more powerful than hand-devised code sharing in the source They are often more powerful than hand-devised code sharing in the source
code. They could be computed off-line by replacing each identifier by code. They could be computed off-line by replacing each identifier by
its definition. its definition.
</P> </P>
<P> <P>
Prefix-suffix tables <B>Prefix-suffix tables</B>
</P> </P>
<PRE> <PRE>
Term ::= "(" String "+" Term ")" ; Term ::= "(" String "+" Term ")" ;
@@ -428,56 +478,6 @@ explains the used syntax rather than the more accurate
since we want the suffix part to be a <CODE>Term</CODE> for the optimization to since we want the suffix part to be a <CODE>Term</CODE> for the optimization to
take effect. take effect.
</P> </P>
<P>
The most curious construct of GFCC is the parameter array alias,
</P>
<PRE>
Term ::= "(" Term "@" Term ")";
</PRE>
<P>
This form is used as the value of parameter records, such as the type
</P>
<PRE>
{n : Number ; p : Person}
</PRE>
<P>
The problem with parameter records is their double role.
They can be used like parameter values, as indices in selection,
</P>
<PRE>
VP.s ! {n = Sg ; p = P3}
</PRE>
<P>
but also as records, from which parameters can be projected:
</P>
<PRE>
{n = Sg ; p = P3}.n
</PRE>
<P>
Whichever use is selected as primary, a prohibitively complex
case expression must be generated at compilation to GFCC to get the
other use. The adopted
solution is to generate a pair containing both a parameter value index
and an array of indices of record fields. For instance, if we have
</P>
<PRE>
param Number = Sg | Pl ; Person = P1 | P2 | P3 ;
</PRE>
<P>
we get the encoding
</P>
<PRE>
{n = Sg ; p = P3} ---&gt; (2 @ [0,2])
</PRE>
<P>
The GFCC computation rules are essentially
</P>
<PRE>
(t ! (i @ _)) = (t ! i)
((_ @ r) ! j) =(r ! j)
</PRE>
<P></P>
<A NAME="toc11"></A>
<H2>Compiling to GFCC</H2> <H2>Compiling to GFCC</H2>
<P> <P>
Compilation to GFCC is performed by the GF grammar compiler, and Compilation to GFCC is performed by the GF grammar compiler, and
@@ -489,32 +489,24 @@ in the process.
The compilation phases are the following The compilation phases are the following
</P> </P>
<OL> <OL>
<LI>translate GF source to GFC, as always in GF <LI>type check and partially evaluate GF source
<LI>undo GFC back-end optimizations <LI>create a symbol table mapping the GF parameter and record types to
<LI>perform the <CODE>values</CODE> optimization to normalize tables
<LI>create a symbol table mapping the GFC parameter and record types to
fixed-size arrays, and parameter values and record labels to integers fixed-size arrays, and parameter values and record labels to integers
<LI>traverse the linearization rules replacing parameters and labels by integers <LI>traverse the linearization rules replacing parameters and labels by integers
<LI>reorganize the created GFC grammar so that it has just one abstract syntax <LI>reorganize the created GF grammar so that it has just one abstract syntax
and one concrete syntax per language and one concrete syntax per language
<LI>apply UTF8 encoding to the grammar, if not yet applied (this is told by the <LI>TODO: apply UTF8 encoding to the grammar, if not yet applied (this is told by the
<CODE>coding</CODE> flag) <CODE>coding</CODE> flag)
<LI>translate the GFC syntax tree to a GFCC syntax tree, using a simple <LI>translate the GF grammar object to a GFCC grammar object, using a simple
compositional mapping compositional mapping
<LI>perform the word-suffix optimization on GFCC linearization terms <LI>perform the word-suffix optimization on GFCC linearization terms
<LI>perform subexpression elimination on each concrete syntax module <LI>perform subexpression elimination on each concrete syntax module
<LI>print out the GFCC code <LI>print out the GFCC code
</OL> </OL>
<P>
Notice that a major part of the compilation is done within GFC, so that
GFC-related tasks (such as parser generation) could be performed by
using the old algorithms.
</P>
<A NAME="toc12"></A>
<H3>Problems in GFCC compilation</H3> <H3>Problems in GFCC compilation</H3>
<P> <P>
Two major problems had to be solved in compiling GFC to GFCC: Two major problems had to be solved in compiling GF to GFCC:
</P> </P>
<UL> <UL>
<LI>consistent order of tables and records, to permit the array translation <LI>consistent order of tables and records, to permit the array translation
@@ -527,17 +519,11 @@ to generate correct code. Any errors remaining are likely to be
related to the two problems just mentioned. related to the two problems just mentioned.
</P> </P>
<P> <P>
The order problem is solved in different ways for tables and records. The order problem is solved in slightly different ways for tables and records.
For tables, the <CODE>values</CODE> optimization of GFC already manages to In both cases, <B>eta expansion</B> is used to establish a
maintain a canonical order. But this order can be destroyed by the canonical order. Tables are ordered by applying the preorder induced
<CODE>share</CODE> optimization. To make sure that GFCC compilation works properly, by <CODE>param</CODE> definitions. Records are ordered by sorting them by labels.
it is safest to recompile the GF grammar by using the <CODE>values</CODE> This means that
optimization flag.
</P>
<P>
Records can be canonically ordered by sorting them by labels.
In fact, this was done in connection of the GFCC work as a part
of the GFC generation, to guarantee consistency. This means that
e.g. the <CODE>s</CODE> field will in general no longer appear as the first e.g. the <CODE>s</CODE> field will in general no longer appear as the first
field, even if it does so in the GF source code. But relying on the field, even if it does so in the GF source code. But relying on the
order of fields in a labelled record would be misplaced anyway. order of fields in a labelled record would be misplaced anyway.
@@ -547,7 +533,7 @@ The canonical form of records is further complicated by lock fields,
i.e. dummy fields of form <CODE>lock_C = &lt;&gt;</CODE>, which are added to grammar i.e. dummy fields of form <CODE>lock_C = &lt;&gt;</CODE>, which are added to grammar
libraries to force intensionality of linearization types. The problem libraries to force intensionality of linearization types. The problem
is that the absence of a lock field only generates a warning, not is that the absence of a lock field only generates a warning, not
an error. Therefore a GFC grammar can contain objects of the same an error. Therefore a GF grammar can contain objects of the same
type with and without a lock field. This problem was solved in GFCC type with and without a lock field. This problem was solved in GFCC
generation by just removing all lock fields (defined as fields whose generation by just removing all lock fields (defined as fields whose
type is the empty record type). This has the further advantage of type is the empty record type). This has the further advantage of
@@ -634,10 +620,22 @@ a case expression,
} }
</PRE> </PRE>
<P> <P>
To avoid the code bloat resulting from this, we chose the alias representation To avoid the code bloat resulting from this, we have chosen to
which is easy enough to deal with in interpreters. deal with records by a <B>currying</B> transformation:
</P> </P>
<A NAME="toc13"></A> <PRE>
table {n : Number ; p : Person} {... ...}
===&gt;
table Number {Sg =&gt; table Person {...} ; table Person {...}}
</PRE>
<P>
This is performed when GFCC is generated. Selections with
records have to be treated likewise,
</P>
<PRE>
t ! r ===&gt; t ! r.n ! r.p
</PRE>
<P></P>
<H3>The representation of linearization types</H3> <H3>The representation of linearization types</H3>
<P> <P>
Linearization types (<CODE>lincat</CODE>) are not needed when generating with Linearization types (<CODE>lincat</CODE>) are not needed when generating with
@@ -647,14 +645,12 @@ concrete syntax, by using terms to represent types. Here is the table
showing how different linearization types are encoded. showing how different linearization types are encoded.
</P> </P>
<PRE> <PRE>
P* = size(P) -- parameter type P* = max(P) -- parameter type
{_ : I ; __ : R}* = (I* @ R*) -- record of parameters {r1 : T1 ; ... ; rn : Tn}* = [T1*,...,Tn*] -- record
{r1 : T1 ; ... ; rn : Tn}* = [T1*,...,Tn*] -- other record (P =&gt; T)* = [T* ,...,T*] -- table, size(P) cases
(P =&gt; T)* = [T* ,...,T*] -- size(P) times
Str* = () Str* = ()
</PRE> </PRE>
<P> <P>
The category symbols are prefixed with two underscores (<CODE>__</CODE>).
For example, the linearization type <CODE>present/CatEng.NP</CODE> is For example, the linearization type <CODE>present/CatEng.NP</CODE> is
translated as follows: translated as follows:
</P> </P>
@@ -667,10 +663,9 @@ translated as follows:
s : {ResEng.Case} =&gt; Str -- 3 values s : {ResEng.Case} =&gt; Str -- 3 values
} }
__NP = [(6@[2,3]),[(),(),()]] __NP = [[1,2],[(),(),()]]
</PRE> </PRE>
<P></P> <P></P>
<A NAME="toc14"></A>
<H3>Running the compiler and the GFCC interpreter</H3> <H3>Running the compiler and the GFCC interpreter</H3>
<P> <P>
GFCC generation is a part of the GFCC generation is a part of the
@@ -679,8 +674,7 @@ of GF since September 2006. To invoke the compiler, the flag
<CODE>-printer=gfcc</CODE> to the command <CODE>-printer=gfcc</CODE> to the command
<CODE>pm = print_multi</CODE> is used. It is wise to recompile the grammar from <CODE>pm = print_multi</CODE> is used. It is wise to recompile the grammar from
source, since previously compiled libraries may not obey the canonical source, since previously compiled libraries may not obey the canonical
order of records. To <CODE>strip</CODE> the grammar before order of records.
GFCC translation removes unnecessary interface references.
Here is an example, performed in Here is an example, performed in
<A HREF="../../../../../examples/bronzeage">example/bronzeage</A>. <A HREF="../../../../../examples/bronzeage">example/bronzeage</A>.
</P> </P>
@@ -690,8 +684,20 @@ Here is an example, performed in
strip strip
pm -printer=gfcc | wf bronze.gfcc pm -printer=gfcc | wf bronze.gfcc
</PRE> </PRE>
<P>
There is also an experimental batch compiler, which does not use the GFC
format or the record aliases. It can be produced by
</P>
<PRE>
make gfc
</PRE>
<P>
in <CODE>GF/src</CODE>, and invoked by
</P>
<PRE>
gfc --make FILES
</PRE>
<P></P> <P></P>
<A NAME="toc15"></A>
<H2>The reference interpreter</H2> <H2>The reference interpreter</H2>
<P> <P>
The reference interpreter written in Haskell consists of the following files: The reference interpreter written in Haskell consists of the following files:
@@ -701,23 +707,37 @@ The reference interpreter written in Haskell consists of the following files:
GFCC.cf -- labelled BNF grammar of gfcc GFCC.cf -- labelled BNF grammar of gfcc
-- files generated by BNFC -- files generated by BNFC
AbsGFCC.hs -- abstrac syntax of gfcc AbsGFCC.hs -- abstrac syntax datatypes
ErrM.hs -- error monad used internally ErrM.hs -- error monad used internally
LexGFCC.hs -- lexer of gfcc files LexGFCC.hs -- lexer of gfcc files
ParGFCC.hs -- parser of gfcc files and syntax trees ParGFCC.hs -- parser of gfcc files and syntax trees
PrintGFCC.hs -- printer of gfcc files and syntax trees PrintGFCC.hs -- printer of gfcc files and syntax trees
-- hand-written files -- hand-written files
DataGFCC.hs -- post-parser grammar creation, linearization and evaluation DataGFCC.hs -- grammar datatype, post-parser grammar creation
GenGFCC.hs -- random and exhaustive generation, generate-and-test parsing Linearize.hs -- linearization and evaluation
RunGFCC.hs -- main function - a simple command interpreter Macros.hs -- utilities abstracting away from GFCC datatypes
Generate.hs -- random and exhaustive generation, generate-and-test parsing
API.hs -- functionalities accessible in embedded GF applications
Generate.hs -- random and exhaustive generation
Shell.hs -- main function - a simple command interpreter
</PRE> </PRE>
<P> <P>
It is included in the It is included in the
<A HREF="http://www.cs.chalmers.se/Cs/Research/Language-technology/darcs/GF/doc/darcs.html">developers' version</A> <A HREF="http://www.cs.chalmers.se/Cs/Research/Language-technology/darcs/GF/doc/darcs.html">developers' version</A>
of GF, in the subdirectory <A HREF="../"><CODE>GF/src/GF/Canon/GFCC</CODE></A>. of GF, in the subdirectories <A HREF="../"><CODE>GF/src/GF/GFCC</CODE></A> and
<A HREF="../../Devel"><CODE>GF/src/GF/Devel</CODE></A>.
</P> </P>
<P> <P>
As of September 2007, default parsing in main GF uses GFCC (implemented by Krasimir
Angelov). The interpreter uses the relevant modules
</P>
<PRE>
GF/Conversions/SimpleToFCFG.hs -- generate parser from GFCC
GF/Parsing/FCFG.hs -- run the parser
</PRE>
<P></P>
<P>
To compile the interpreter, type To compile the interpreter, type
</P> </P>
<PRE> <PRE>
@@ -741,87 +761,34 @@ The available commands are
and show their linearizations in all languages and show their linearizations in all languages
<LI><CODE>gtt &lt;Cat&gt; &lt;Int&gt;</CODE>: generate a number of trees in category from smallest, <LI><CODE>gtt &lt;Cat&gt; &lt;Int&gt;</CODE>: generate a number of trees in category from smallest,
and show the trees and their linearizations in all languages and show the trees and their linearizations in all languages
<LI><CODE>p &lt;Int&gt; &lt;Cat&gt; &lt;String&gt;</CODE>: "parse", i.e. generate trees until match or <LI><CODE>p &lt;Lang&gt; &lt;Cat&gt; &lt;String&gt;</CODE>: parse a string into a set of trees
until the given number have been generated <LI><CODE>lin &lt;Tree&gt;</CODE>: linearize tree in all languages, also showing full records
<LI><CODE>&lt;Tree&gt;</CODE>: linearize tree in all languages, also showing full records <LI><CODE>q</CODE>: terminate the system cleanly
<LI><CODE>quit</CODE>: terminate the system cleanly
</UL> </UL>
<A NAME="toc16"></A> <H2>Embedded formats</H2>
<H2>Interpreter in C++</H2>
<P>
A base-line interpreter in C++ has been started.
Its main functionality is random generation of trees and linearization of them.
</P>
<P>
Here are some results from running the different interpreters, compared
to running the same grammar in GF, saved in <CODE>.gfcm</CODE> format.
The grammar contains the English, German, and Norwegian
versions of Bronzeage. The experiment was carried out on
Ubuntu Linux laptop with 1.5 GHz Intel centrino processor.
</P>
<TABLE CELLPADDING="4" BORDER="1">
<TR>
<TH></TH>
<TH>GF</TH>
<TH>gfcc(hs)</TH>
<TH>gfcc++</TH>
</TR>
<TR>
<TD>program size</TD>
<TD ALIGN="center">7249k</TD>
<TD ALIGN="center">803k</TD>
<TD ALIGN="right">113k</TD>
</TR>
<TR>
<TD>grammar size</TD>
<TD ALIGN="center">336k</TD>
<TD ALIGN="center">119k</TD>
<TD ALIGN="right">119k</TD>
</TR>
<TR>
<TD>read grammar</TD>
<TD ALIGN="center">1150ms</TD>
<TD ALIGN="center">510ms</TD>
<TD ALIGN="right">100ms</TD>
</TR>
<TR>
<TD>generate 222</TD>
<TD ALIGN="center">9500ms</TD>
<TD ALIGN="center">450ms</TD>
<TD ALIGN="right">800ms</TD>
</TR>
<TR>
<TD>memory</TD>
<TD ALIGN="center">21M</TD>
<TD ALIGN="center">10M</TD>
<TD ALIGN="right">20M</TD>
</TR>
</TABLE>
<P></P>
<P>
To summarize:
</P>
<UL> <UL>
<LI>going from GF to gfcc is a major win in both code size and efficiency <LI>JavaScript: compiler of linearization and abstract syntax
<LI>going from Haskell to C++ interpreter is not a win yet, because of a space <P></P>
leak in the C++ version <LI>Haskell: compiler of abstract syntax and interpreter with parsing,
linearization, and generation
<P></P>
<LI>C: compiler of linearization (old GFCC)
<P></P>
<LI>C++: embedded interpreter supporting linearization (old GFCC)
</UL> </UL>
<A NAME="toc17"></A>
<H2>Some things to do</H2> <H2>Some things to do</H2>
<P> <P>
Support for dependent types, higher-order abstract syntax, and
semantic definition in GFCC generation and interpreters.
</P>
<P>
Replacing the entire GF shell by one based on GFCC.
</P>
<P>
Interpreter in Java. Interpreter in Java.
</P> </P>
<P>
Parsing via MCFG
</P>
<UL>
<LI>the FCFG format can possibly be simplified
<LI>parser grammars should be saved in files to make interpreters easier
</UL>
<P> <P>
Hand-written parsers for GFCC grammars to reduce code size Hand-written parsers for GFCC grammars to reduce code size
(and efficiency?) of interpreters. (and efficiency?) of interpreters.
@@ -838,5 +805,5 @@ word-suffix sharing better (depth-one tables, as in FM).
</P> </P>
<!-- html code generated by txt2tags 2.3 (http://txt2tags.sf.net) --> <!-- html code generated by txt2tags 2.3 (http://txt2tags.sf.net) -->
<!-- cmdline: txt2tags -thtml -\-toc gfcc.txt --> <!-- cmdline: txt2tags -thtml gfcc.txt -->
</BODY></HTML> </BODY></HTML>

430
src/GF/GFCC/doc/gfcc.txt
View File

@@ -1,6 +1,6 @@
The GFCC Grammar Format The GFCC Grammar Format
Aarne Ranta Aarne Ranta
October 19, 2006 October 5, 2007
Author's address: Author's address:
[``http://www.cs.chalmers.se/~aarne`` http://www.cs.chalmers.se/~aarne] [``http://www.cs.chalmers.se/~aarne`` http://www.cs.chalmers.se/~aarne]
@@ -8,6 +8,7 @@ Author's address:
% to compile: txt2tags -thtml --toc gfcc.txt % to compile: txt2tags -thtml --toc gfcc.txt
History: History:
- 5 Oct 2007: new, better structured GFCC with full expressive power
- 19 Oct: translation of lincats, new figures on C++ - 19 Oct: translation of lincats, new figures on C++
- 3 Oct 2006: first version - 3 Oct 2006: first version
@@ -22,13 +23,15 @@ advantages:
- simple definition of interpreters - simple definition of interpreters
The idea is that all embedded GF applications are compiled to GFCC. Thus we also want to call GFCC the **portable grammar format**.
The idea is that all embedded GF applications use GFCC.
The GF system would be primarily used as a compiler and as a grammar The GF system would be primarily used as a compiler and as a grammar
development tool. development tool.
Since GFCC is implemented in BNFC, a parser of the format is readily Since GFCC is implemented in BNFC, a parser of the format is readily
available for C, C++, Haskell, Java, and OCaml. Also an XML available for C, C++, C#, Haskell, Java, and OCaml. Also an XML
representation is generated in BNFC. A representation can be generated in BNFC. A
[reference implementation ../] [reference implementation ../]
of linearization and some other functions has been written in Haskell. of linearization and some other functions has been written in Haskell.
@@ -44,61 +47,68 @@ such as type annotations, which is only needed in compilation and not at
run-time. In particular, the pattern matching syntax and semantics of GFC is run-time. In particular, the pattern matching syntax and semantics of GFC is
complex and therefore difficult to implement in new platforms. complex and therefore difficult to implement in new platforms.
The main differences of GFCC compared with GFC can be summarized as follows: Actually, GFC is planned to be omitted also as the target format of
separate compilation, where plain GF (type annotated and partially evaluated)
will be used instead. GFC provides only marginal advantages as a target format
compared with GF, and it is therefore just extra weight to carry around this
format.
The main differences of GFCC compared with GFC (and GF) can be summarized as follows:
- there are no modules, and therefore no qualified names - there are no modules, and therefore no qualified names
- a GFCC grammar is multilingual, and consists of a common abstract syntax - a GFCC grammar is multilingual, and consists of a common abstract syntax
together with one concrete syntax per language together with one concrete syntax per language
- records and tables are replaced by arrays - records and tables are replaced by arrays
- record labels and parameter values are replaced by integers - record labels and parameter values are replaced by integers
- record projection and table selection are replaced by array indexing - record projection and table selection are replaced by array indexing
- there is (so far) no support for dependent types or higher-order abstract - even though the format does support dependent types and higher-order abstract
syntax (which would be easy to add, but make interpreters much more difficult syntax, there is no interpreted yet that does this
to write)
Here is an example of a GF grammar, consisting of three modules, Here is an example of a GF grammar, consisting of three modules,
as translated to GFCC. The representations are aligned, with the exceptions as translated to GFCC. The representations are aligned; thus they do not completely
due to the alphabetical sorting of GFCC grammars. reflect the order of judgements in GFCC files, which have different orders of
blocks of judgements, and alphabetical sorting.
``` ```
grammar Ex(Eng,Swe); grammar Ex(Eng,Swe);
abstract Ex = { abstract { abstract Ex = { abstract {
cat cat cat
S ; NP ; VP ; S ; NP ; VP ; NP[]; S[]; VP[];
fun fun fun
Pred : NP -> VP -> S ; Pred : NP,VP -> S = (Pred); Pred : NP -> VP -> S ; Pred=[(($ 0! 1),(($ 1! 0)!($ 0! 0)))];
She, They : NP ; She : -> NP = (She); She, They : NP ; She=[0,"she"];
Sleep : VP ; Sleep : -> VP = (Sleep); Sleep : VP ; They=[1,"they"];
They : -> NP = (They); Sleep=[["sleeps","sleep"]];
} } ; } } ;
concrete Eng of Ex = { concrete Eng { concrete Eng of Ex = { concrete Eng {
lincat lincat lincat
S = {s : Str} ; S = {s : Str} ; S=[()];
NP = {s : Str ; n : Num} ; NP = {s : Str ; n : Num} ; NP=[1,()];
VP = {s : Num => Str} ; VP = {s : Num => Str} ; VP=[[(),()]];
param param
Num = Sg | Pl ; Num = Sg | Pl ;
lin lin lin
Pred np vp = { Pred = [(($0!1),(($1!0)!($0!0)))]; Pred np vp = { Pred=[(($ 0! 1),(($ 1! 0)!($ 0! 0)))];
s = np.s ++ vp.s ! np.n} ; s = np.s ++ vp.s ! np.n} ;
She = {s = "she" ; n = Sg} ; She = [0, "she"]; She = {s = "she" ; n = Sg} ; She=[0,"she"];
They = {s = "they" ; n = Pl} ; They = {s = "they" ; n = Pl} ; They = [1, "they"];
Sleep = {s = table { Sleep = [("sleep" + ["s",""])]; Sleep = {s = table { Sleep=[["sleeps","sleep"]];
Sg => "sleeps" ; Sg => "sleeps" ;
Pl => "sleep" They = [1, "they"]; Pl => "sleep"
} } ; }
} ; } ;
} } } ;
concrete Swe of Ex = { concrete Swe { concrete Swe of Ex = { concrete Swe {
lincat lincat lincat
S = {s : Str} ; S = {s : Str} ; S=[()];
NP = {s : Str} ; NP = {s : Str} ; NP=[()];
VP = {s : Str} ; VP = {s : Str} ; VP=[()];
param param
Num = Sg | Pl ; Num = Sg | Pl ;
lin lin lin
Pred np vp = { Pred = [(($0!0),($1!0))]; Pred np vp = { Pred = [(($0!0),($1!0))];
s = np.s ++ vp.s} ; s = np.s ++ vp.s} ;
She = {s = "hon"} ; She = ["hon"]; She = {s = "hon"} ; She = ["hon"];
@@ -109,6 +119,11 @@ concrete Swe of Ex = { concrete Swe {
==The syntax of GFCC files== ==The syntax of GFCC files==
The complete BNFC grammar, from which
the rules in this section are taken, is in the file
[``GF/GFCC/GFCC.cf`` ../DataGFCC.cf].
===Top level=== ===Top level===
A grammar has a header telling the name of the abstract syntax A grammar has a header telling the name of the abstract syntax
@@ -116,21 +131,40 @@ A grammar has a header telling the name of the abstract syntax
the concrete languages. The abstract syntax and the concrete the concrete languages. The abstract syntax and the concrete
syntaxes themselves follow. syntaxes themselves follow.
``` ```
Grammar ::= Header ";" Abstract ";" [Concrete] ; Grm. Grammar ::=
Header ::= "grammar" CId "(" [CId] ")" ; "grammar" CId "(" [CId] ")" ";"
Abstract ::= "abstract" "{" [AbsDef] "}" ; Abstract ";"
Concrete ::= "concrete" CId "{" [CncDef] "}" ; [Concrete] ;
Abs. Abstract ::=
"abstract" "{"
"flags" [Flag]
"fun" [FunDef]
"cat" [CatDef]
"}" ;
Cnc. Concrete ::=
"concrete" CId "{"
"flags" [Flag]
"lin" [LinDef]
"oper" [LinDef]
"lincat" [LinDef]
"lindef" [LinDef]
"printname" [LinDef]
"}" ;
``` ```
Abstract syntax judgements give typings and semantic definitions. This syntax organizes each module to a sequence of **fields**, such
Concrete syntax judgements give linearizations. as flags, linearizations, operations, linearization types, etc.
It is envisaged that particular applications can ignore some
of the fields, typically so that earlier fields are more
important than later ones.
The judgement forms have the following syntax.
``` ```
AbsDef ::= CId ":" Type "=" Exp ; Flg. Flag ::= CId "=" String ;
CncDef ::= CId "=" Term ; Cat. CatDef ::= CId "[" [Hypo] "]" ;
``` Fun. FunDef ::= CId ":" Type "=" Exp ;
Also flags are possible, local to each "module" (i.e. abstract and concretes). Lin. LinDef ::= CId "=" Term ;
```
AbsDef ::= "%" CId "=" String ;
CncDef ::= "%" CId "=" String ;
``` ```
For the run-time system, the reference implementation in Haskell For the run-time system, the reference implementation in Haskell
uses a structure that gives efficient look-up: uses a structure that gives efficient look-up:
@@ -143,30 +177,74 @@ uses a structure that gives efficient look-up:
} }
data Abstr = Abstr { data Abstr = Abstr {
funs :: Map CId Type, -- find the type of a fun aflags :: Map CId String, -- value of a flag
cats :: Map CId [CId] -- find the funs giving a cat funs :: Map CId (Type,Exp), -- type and def of a fun
cats :: Map CId [Hypo], -- context of a cat
catfuns :: Map CId [CId] -- funs yielding a cat (redundant, for fast lookup)
} }
type Concr = Map CId Term data Concr = Concr {
flags :: Map CId String, -- value of a flag
lins :: Map CId Term, -- lin of a fun
opers :: Map CId Term, -- oper generated by subex elim
lincats :: Map CId Term, -- lin type of a cat
lindefs :: Map CId Term, -- lin default of a cat
printnames :: Map CId Term -- printname of a cat or a fun
}
```
These definitions are from [``GF/GFCC/DataGFCC.hs`` ../DataGFCC.hs].
Identifiers (``CId``) are like ``Ident`` in GF, except that
the compiler produces constants prefixed with ``_`` in
the common subterm elimination optimization.
```
token CId (('_' | letter) (letter | digit | '\'' | '_')*) ;
``` ```
===Abstract syntax=== ===Abstract syntax===
Types are first-order function types built from Types are first-order function types built from argument type
contexts and value types.
category symbols. Syntax trees (``Exp``) are category symbols. Syntax trees (``Exp``) are
rose trees with the head (``Atom``) either a function rose trees with nodes consisting of a head (``Atom``) and
constant, a metavariable, or a string, integer, or float bound variables (``CId``).
```
DTyp. Type ::= "[" [Hypo] "]" CId [Exp] ;
DTr. Exp ::= "[" "(" [CId] ")" Atom [Exp] "]" ;
Hyp. Hypo ::= CId ":" Type ;
```
The head Atom is either a function
constant, a bound variable, or a metavariable, or a string, integer, or float
literal. literal.
``` ```
Type ::= [CId] "->" CId ; AC. Atom ::= CId ;
Exp ::= "(" Atom [Exp] ")" ; AS. Atom ::= String ;
Atom ::= CId ; -- function constant AI. Atom ::= Integer ;
Atom ::= "?" ; -- metavariable AF. Atom ::= Double ;
Atom ::= String ; -- string literal AM. Atom ::= "?" Integer ;
Atom ::= Integer ; -- integer literal
Atom ::= Double ; -- float literal
``` ```
The context-free types and trees of the "old GFCC" are special
cases, which can be defined as follows:
```
Typ. Type ::= [CId] "->" CId
Typ args val = DTyp [Hyp (CId "_") arg | arg <- args] val
Tr. Exp ::= "(" CId [Exp] ")"
Tr fun exps = DTr [] fun exps
```
To store semantic (``def``) definitions by cases, the following expression
form is provided, but it is only meaningful in the last field of a function
declaration in an abstract syntax:
```
EEq. Exp ::= "{" [Equation] "}" ;
Equ. Equation ::= [Exp] "->" Exp ;
```
Notice that expressions are used to encode patterns. Primitive notions
(the default semantics in GF) are encoded as empty sets of equations
(``[]``). For a constructor (canonical form) of a category ``C``, we
aim to use the encoding as the application ``(_constr C)``.
===Concrete syntax=== ===Concrete syntax===
@@ -175,12 +253,12 @@ Linearization terms (``Term``) are built as follows.
Constructor names are shown to make the later code Constructor names are shown to make the later code
examples readable. examples readable.
``` ```
R. Term ::= "[" [Term] "]" ; -- array R. Term ::= "[" [Term] "]" ; -- array (record/table)
P. Term ::= "(" Term "!" Term ")" ; -- access to indexed field P. Term ::= "(" Term "!" Term ")" ; -- access to field (projection/selection)
S. Term ::= "(" [Term] ")" ; -- sequence with ++ S. Term ::= "(" [Term] ")" ; -- concatenated sequence
K. Term ::= Tokn ; -- token K. Term ::= Tokn ; -- token
V. Term ::= "$" Integer ; -- argument V. Term ::= "$" Integer ; -- argument (subtree)
C. Term ::= Integer ; -- array index C. Term ::= Integer ; -- array index (label/parameter value)
FV. Term ::= "[|" [Term] "|]" ; -- free variation FV. Term ::= "[|" [Term] "|]" ; -- free variation
TM. Term ::= "?" ; -- linearization of metavariable TM. Term ::= "?" ; -- linearization of metavariable
``` ```
@@ -191,25 +269,27 @@ variant lists.
KP. Tokn ::= "[" "pre" [String] "[" [Variant] "]" "]" ; KP. Tokn ::= "[" "pre" [String] "[" [Variant] "]" "]" ;
Var. Variant ::= [String] "/" [String] ; Var. Variant ::= [String] "/" [String] ;
``` ```
Three special forms of terms are introduced by the compiler Two special forms of terms are introduced by the compiler
as optimizations. They can in principle be eliminated, but as optimizations. They can in principle be eliminated, but
their presence makes grammars much more compact. Their semantics their presence makes grammars much more compact. Their semantics
will be explained in a later section. will be explained in a later section.
``` ```
F. Term ::= CId ; -- global constant F. Term ::= CId ; -- global constant
W. Term ::= "(" String "+" Term ")" ; -- prefix + suffix table W. Term ::= "(" String "+" Term ")" ; -- prefix + suffix table
RP. Term ::= "(" Term "@" Term ")"; -- record parameter alias
``` ```
Identifiers are like ``Ident`` in GF and GFC, except that There is also a deprecated form of "record parameter alias",
the compiler produces constants prefixed with ``_`` in
the common subterm elimination optimization.
``` ```
token CId (('_' | letter) (letter | digit | '\'' | '_')*) ; RP. Term ::= "(" Term "@" Term ")"; -- DEPRECATED
``` ```
which will be removed when the migration to new GFCC is complete.
==The semantics of concrete syntax terms== ==The semantics of concrete syntax terms==
The code in this section is from [``GF/GFCC/Linearize.hs`` ../Linearize.hs].
===Linearization and realization=== ===Linearization and realization===
The linearization algorithm is essentially the same as in The linearization algorithm is essentially the same as in
@@ -220,17 +300,19 @@ The function also needs to know the language (i.e. concrete syntax)
in which linearization is performed. in which linearization is performed.
``` ```
linExp :: GFCC -> CId -> Exp -> Term linExp :: GFCC -> CId -> Exp -> Term
linExp mcfg lang tree@(Tr at trees) = case at of linExp gfcc lang tree@(DTr _ at trees) = case at of
AC fun -> comp (Prelude.map lin trees) $ look fun AC fun -> comp (Prelude.map lin trees) $ look fun
AS s -> R [kks (show s)] -- quoted AS s -> R [kks (show s)] -- quoted
AI i -> R [kks (show i)] AI i -> R [kks (show i)]
AF d -> R [kks (show d)] AF d -> R [kks (show d)]
AM -> TM AM -> TM
where where
lin = linExp mcfg lang lin = linExp gfcc lang
comp = compute mcfg lang comp = compute gfcc lang
look = lookLin mcfg lang look = lookLin gfcc lang
``` ```
TODO: bindings must be supported.
The result of linearization is usually a record, which is realized as The result of linearization is usually a record, which is realized as
a string using the following algorithm. a string using the following algorithm.
``` ```
@@ -244,10 +326,11 @@ a string using the following algorithm.
FV (t:_) -> realize t FV (t:_) -> realize t
TM -> "?" TM -> "?"
``` ```
Since the order of record fields is not necessarily Notice that realization always picks the first field of a record.
the same as in GF source, If a linearization type has more than one field, the first field
this realization does not work securely for does not necessarily contain the desired string.
categories whose lincats more than one field. Also notice that the order of record fields in GFCC is not necessarily
the same as in GF source.
===Term evaluation=== ===Term evaluation===
@@ -263,10 +346,9 @@ enable reimplementations in languages that do not permit
deep patterns (such as Java and C++). deep patterns (such as Java and C++).
``` ```
compute :: GFCC -> CId -> [Term] -> Term -> Term compute :: GFCC -> CId -> [Term] -> Term -> Term
compute mcfg lang args = comp where compute gfcc lang args = comp where
comp trm = case trm of comp trm = case trm of
P r p -> proj (comp r) (comp p) P r p -> proj (comp r) (comp p)
RP i t -> RP (comp i) (comp t)
W s t -> W s (comp t) W s t -> W s (comp t)
R ts -> R $ Prelude.map comp ts R ts -> R $ Prelude.map comp ts
V i -> idx args (fromInteger i) -- already computed V i -> idx args (fromInteger i) -- already computed
@@ -275,7 +357,7 @@ compute mcfg lang args = comp where
S ts -> S $ Prelude.filter (/= S []) $ Prelude.map comp ts S ts -> S $ Prelude.filter (/= S []) $ Prelude.map comp ts
_ -> trm _ -> trm
look = lookLin mcfg lang look = lookOper gfcc lang
idx xs i = xs !! i idx xs i = xs !! i
@@ -311,12 +393,12 @@ Global constants
Term ::= CId ; Term ::= CId ;
``` ```
are shorthands for complex terms. They are produced by the are shorthands for complex terms. They are produced by the
compiler by (iterated) common subexpression elimination. compiler by (iterated) **common subexpression elimination**.
They are often more powerful than hand-devised code sharing in the source They are often more powerful than hand-devised code sharing in the source
code. They could be computed off-line by replacing each identifier by code. They could be computed off-line by replacing each identifier by
its definition. its definition.
Prefix-suffix tables **Prefix-suffix tables**
``` ```
Term ::= "(" String "+" Term ")" ; Term ::= "(" String "+" Term ")" ;
``` ```
@@ -339,40 +421,6 @@ explains the used syntax rather than the more accurate
since we want the suffix part to be a ``Term`` for the optimization to since we want the suffix part to be a ``Term`` for the optimization to
take effect. take effect.
The most curious construct of GFCC is the parameter array alias,
```
Term ::= "(" Term "@" Term ")";
```
This form is used as the value of parameter records, such as the type
```
{n : Number ; p : Person}
```
The problem with parameter records is their double role.
They can be used like parameter values, as indices in selection,
```
VP.s ! {n = Sg ; p = P3}
```
but also as records, from which parameters can be projected:
```
{n = Sg ; p = P3}.n
```
Whichever use is selected as primary, a prohibitively complex
case expression must be generated at compilation to GFCC to get the
other use. The adopted
solution is to generate a pair containing both a parameter value index
and an array of indices of record fields. For instance, if we have
```
param Number = Sg | Pl ; Person = P1 | P2 | P3 ;
```
we get the encoding
```
{n = Sg ; p = P3} ---> (2 @ [0,2])
```
The GFCC computation rules are essentially
```
(t ! (i @ _)) = (t ! i)
((_ @ r) ! j) =(r ! j)
```
==Compiling to GFCC== ==Compiling to GFCC==
@@ -383,31 +431,26 @@ however, it might be interesting to know what happens to the grammars
in the process. in the process.
The compilation phases are the following The compilation phases are the following
+ translate GF source to GFC, as always in GF + type check and partially evaluate GF source
+ undo GFC back-end optimizations + create a symbol table mapping the GF parameter and record types to
+ perform the ``values`` optimization to normalize tables
+ create a symbol table mapping the GFC parameter and record types to
fixed-size arrays, and parameter values and record labels to integers fixed-size arrays, and parameter values and record labels to integers
+ traverse the linearization rules replacing parameters and labels by integers + traverse the linearization rules replacing parameters and labels by integers
+ reorganize the created GFC grammar so that it has just one abstract syntax + reorganize the created GF grammar so that it has just one abstract syntax
and one concrete syntax per language and one concrete syntax per language
+ apply UTF8 encoding to the grammar, if not yet applied (this is told by the + TODO: apply UTF8 encoding to the grammar, if not yet applied (this is told by the
``coding`` flag) ``coding`` flag)
+ translate the GFC syntax tree to a GFCC syntax tree, using a simple + translate the GF grammar object to a GFCC grammar object, using a simple
compositional mapping compositional mapping
+ perform the word-suffix optimization on GFCC linearization terms + perform the word-suffix optimization on GFCC linearization terms
+ perform subexpression elimination on each concrete syntax module + perform subexpression elimination on each concrete syntax module
+ print out the GFCC code + print out the GFCC code
Notice that a major part of the compilation is done within GFC, so that
GFC-related tasks (such as parser generation) could be performed by
using the old algorithms.
===Problems in GFCC compilation=== ===Problems in GFCC compilation===
Two major problems had to be solved in compiling GFC to GFCC: Two major problems had to be solved in compiling GF to GFCC:
- consistent order of tables and records, to permit the array translation - consistent order of tables and records, to permit the array translation
- run-time variables in complex parameter values. - run-time variables in complex parameter values.
@@ -416,16 +459,11 @@ The current implementation is still experimental and may fail
to generate correct code. Any errors remaining are likely to be to generate correct code. Any errors remaining are likely to be
related to the two problems just mentioned. related to the two problems just mentioned.
The order problem is solved in different ways for tables and records. The order problem is solved in slightly different ways for tables and records.
For tables, the ``values`` optimization of GFC already manages to In both cases, **eta expansion** is used to establish a
maintain a canonical order. But this order can be destroyed by the canonical order. Tables are ordered by applying the preorder induced
``share`` optimization. To make sure that GFCC compilation works properly, by ``param`` definitions. Records are ordered by sorting them by labels.
it is safest to recompile the GF grammar by using the ``values`` This means that
optimization flag.
Records can be canonically ordered by sorting them by labels.
In fact, this was done in connection of the GFCC work as a part
of the GFC generation, to guarantee consistency. This means that
e.g. the ``s`` field will in general no longer appear as the first e.g. the ``s`` field will in general no longer appear as the first
field, even if it does so in the GF source code. But relying on the field, even if it does so in the GF source code. But relying on the
order of fields in a labelled record would be misplaced anyway. order of fields in a labelled record would be misplaced anyway.
@@ -434,7 +472,7 @@ The canonical form of records is further complicated by lock fields,
i.e. dummy fields of form ``lock_C = <>``, which are added to grammar i.e. dummy fields of form ``lock_C = <>``, which are added to grammar
libraries to force intensionality of linearization types. The problem libraries to force intensionality of linearization types. The problem
is that the absence of a lock field only generates a warning, not is that the absence of a lock field only generates a warning, not
an error. Therefore a GFC grammar can contain objects of the same an error. Therefore a GF grammar can contain objects of the same
type with and without a lock field. This problem was solved in GFCC type with and without a lock field. This problem was solved in GFCC
generation by just removing all lock fields (defined as fields whose generation by just removing all lock fields (defined as fields whose
type is the empty record type). This has the further advantage of type is the empty record type). This has the further advantage of
@@ -503,8 +541,18 @@ a case expression,
5 => 1 5 => 1
} }
``` ```
To avoid the code bloat resulting from this, we chose the alias representation To avoid the code bloat resulting from this, we have chosen to
which is easy enough to deal with in interpreters. deal with records by a **currying** transformation:
```
table {n : Number ; p : Person} {... ...}
===>
table Number {Sg => table Person {...} ; table Person {...}}
```
This is performed when GFCC is generated. Selections with
records have to be treated likewise,
```
t ! r ===> t ! r.n ! r.p
```
===The representation of linearization types=== ===The representation of linearization types===
@@ -515,13 +563,11 @@ GFCC. The linearization type definitions are shown as a part of the
concrete syntax, by using terms to represent types. Here is the table concrete syntax, by using terms to represent types. Here is the table
showing how different linearization types are encoded. showing how different linearization types are encoded.
``` ```
P* = size(P) -- parameter type P* = max(P) -- parameter type
{_ : I ; __ : R}* = (I* @ R*) -- record of parameters {r1 : T1 ; ... ; rn : Tn}* = [T1*,...,Tn*] -- record
{r1 : T1 ; ... ; rn : Tn}* = [T1*,...,Tn*] -- other record (P => T)* = [T* ,...,T*] -- table, size(P) cases
(P => T)* = [T* ,...,T*] -- size(P) times
Str* = () Str* = ()
``` ```
The category symbols are prefixed with two underscores (``__``).
For example, the linearization type ``present/CatEng.NP`` is For example, the linearization type ``present/CatEng.NP`` is
translated as follows: translated as follows:
``` ```
@@ -533,7 +579,7 @@ translated as follows:
s : {ResEng.Case} => Str -- 3 values s : {ResEng.Case} => Str -- 3 values
} }
__NP = [(6@[2,3]),[(),(),()]] __NP = [[1,2],[(),(),()]]
``` ```
@@ -547,8 +593,7 @@ of GF since September 2006. To invoke the compiler, the flag
``-printer=gfcc`` to the command ``-printer=gfcc`` to the command
``pm = print_multi`` is used. It is wise to recompile the grammar from ``pm = print_multi`` is used. It is wise to recompile the grammar from
source, since previously compiled libraries may not obey the canonical source, since previously compiled libraries may not obey the canonical
order of records. To ``strip`` the grammar before order of records.
GFCC translation removes unnecessary interface references.
Here is an example, performed in Here is an example, performed in
[example/bronzeage ../../../../../examples/bronzeage]. [example/bronzeage ../../../../../examples/bronzeage].
``` ```
@@ -557,6 +602,16 @@ Here is an example, performed in
strip strip
pm -printer=gfcc | wf bronze.gfcc pm -printer=gfcc | wf bronze.gfcc
``` ```
There is also an experimental batch compiler, which does not use the GFC
format or the record aliases. It can be produced by
```
make gfc
```
in ``GF/src``, and invoked by
```
gfc --make FILES
```
@@ -568,20 +623,33 @@ The reference interpreter written in Haskell consists of the following files:
GFCC.cf -- labelled BNF grammar of gfcc GFCC.cf -- labelled BNF grammar of gfcc
-- files generated by BNFC -- files generated by BNFC
AbsGFCC.hs -- abstrac syntax of gfcc AbsGFCC.hs -- abstrac syntax datatypes
ErrM.hs -- error monad used internally ErrM.hs -- error monad used internally
LexGFCC.hs -- lexer of gfcc files LexGFCC.hs -- lexer of gfcc files
ParGFCC.hs -- parser of gfcc files and syntax trees ParGFCC.hs -- parser of gfcc files and syntax trees
PrintGFCC.hs -- printer of gfcc files and syntax trees PrintGFCC.hs -- printer of gfcc files and syntax trees
-- hand-written files -- hand-written files
DataGFCC.hs -- post-parser grammar creation, linearization and evaluation DataGFCC.hs -- grammar datatype, post-parser grammar creation
GenGFCC.hs -- random and exhaustive generation, generate-and-test parsing Linearize.hs -- linearization and evaluation
RunGFCC.hs -- main function - a simple command interpreter Macros.hs -- utilities abstracting away from GFCC datatypes
Generate.hs -- random and exhaustive generation, generate-and-test parsing
API.hs -- functionalities accessible in embedded GF applications
Generate.hs -- random and exhaustive generation
Shell.hs -- main function - a simple command interpreter
``` ```
It is included in the It is included in the
[developers' version http://www.cs.chalmers.se/Cs/Research/Language-technology/darcs/GF/doc/darcs.html] [developers' version http://www.cs.chalmers.se/Cs/Research/Language-technology/darcs/GF/doc/darcs.html]
of GF, in the subdirectory [``GF/src/GF/Canon/GFCC`` ../]. of GF, in the subdirectories [``GF/src/GF/GFCC`` ../] and
[``GF/src/GF/Devel`` ../../Devel].
As of September 2007, default parsing in main GF uses GFCC (implemented by Krasimir
Angelov). The interpreter uses the relevant modules
```
GF/Conversions/SimpleToFCFG.hs -- generate parser from GFCC
GF/Parsing/FCFG.hs -- run the parser
```
To compile the interpreter, type To compile the interpreter, type
``` ```
@@ -600,48 +668,34 @@ The available commands are
and show their linearizations in all languages and show their linearizations in all languages
- ``gtt <Cat> <Int>``: generate a number of trees in category from smallest, - ``gtt <Cat> <Int>``: generate a number of trees in category from smallest,
and show the trees and their linearizations in all languages and show the trees and their linearizations in all languages
- ``p <Int> <Cat> <String>``: "parse", i.e. generate trees until match or - ``p <Lang> <Cat> <String>``: parse a string into a set of trees
until the given number have been generated - ``lin <Tree>``: linearize tree in all languages, also showing full records
- ``<Tree>``: linearize tree in all languages, also showing full records - ``q``: terminate the system cleanly
- ``quit``: terminate the system cleanly
==Interpreter in C++==
A base-line interpreter in C++ has been started.
Its main functionality is random generation of trees and linearization of them.
Here are some results from running the different interpreters, compared
to running the same grammar in GF, saved in ``.gfcm`` format.
The grammar contains the English, German, and Norwegian
versions of Bronzeage. The experiment was carried out on
Ubuntu Linux laptop with 1.5 GHz Intel centrino processor.
|| | GF | gfcc(hs) | gfcc++ |
| program size | 7249k | 803k | 113k
| grammar size | 336k | 119k | 119k
| read grammar | 1150ms | 510ms | 100ms
| generate 222 | 9500ms | 450ms | 800ms
| memory | 21M | 10M | 20M
To summarize: ==Embedded formats==
- going from GF to gfcc is a major win in both code size and efficiency
- going from Haskell to C++ interpreter is not a win yet, because of a space - JavaScript: compiler of linearization and abstract syntax
leak in the C++ version
- Haskell: compiler of abstract syntax and interpreter with parsing,
linearization, and generation
- C: compiler of linearization (old GFCC)
- C++: embedded interpreter supporting linearization (old GFCC)
==Some things to do== ==Some things to do==
Support for dependent types, higher-order abstract syntax, and
semantic definition in GFCC generation and interpreters.
Replacing the entire GF shell by one based on GFCC.
Interpreter in Java. Interpreter in Java.
Parsing via MCFG
- the FCFG format can possibly be simplified
- parser grammars should be saved in files to make interpreters easier
Hand-written parsers for GFCC grammars to reduce code size Hand-written parsers for GFCC grammars to reduce code size
(and efficiency?) of interpreters. (and efficiency?) of interpreters.
@@ -652,5 +706,3 @@ Syntax editor based on GFCC.
Rewriting of resource libraries in order to exploit the Rewriting of resource libraries in order to exploit the
word-suffix sharing better (depth-one tables, as in FM). word-suffix sharing better (depth-one tables, as in FM).

656
src/GF/GFCC/doc/old-gfcc.txt Normal file
View File

@@ -0,0 +1,656 @@
The GFCC Grammar Format
Aarne Ranta
October 19, 2006
Author's address:
[``http://www.cs.chalmers.se/~aarne`` http://www.cs.chalmers.se/~aarne]
% to compile: txt2tags -thtml --toc gfcc.txt
History:
- 19 Oct: translation of lincats, new figures on C++
- 3 Oct 2006: first version
==What is GFCC==
GFCC is a low-level format for GF grammars. Its aim is to contain the minimum
that is needed to process GF grammars at runtime. This minimality has three
advantages:
- compact grammar files and run-time objects
- time and space efficient processing
- simple definition of interpreters
The idea is that all embedded GF applications are compiled to GFCC.
The GF system would be primarily used as a compiler and as a grammar
development tool.
Since GFCC is implemented in BNFC, a parser of the format is readily
available for C, C++, Haskell, Java, and OCaml. Also an XML
representation is generated in BNFC. A
[reference implementation ../]
of linearization and some other functions has been written in Haskell.
==GFCC vs. GFC==
GFCC is aimed to replace GFC as the run-time grammar format. GFC was designed
to be a run-time format, but also to
support separate compilation of grammars, i.e.
to store the results of compiling
individual GF modules. But this means that GFC has to contain extra information,
such as type annotations, which is only needed in compilation and not at
run-time. In particular, the pattern matching syntax and semantics of GFC is
complex and therefore difficult to implement in new platforms.
The main differences of GFCC compared with GFC can be summarized as follows:
- there are no modules, and therefore no qualified names
- a GFCC grammar is multilingual, and consists of a common abstract syntax
together with one concrete syntax per language
- records and tables are replaced by arrays
- record labels and parameter values are replaced by integers
- record projection and table selection are replaced by array indexing
- there is (so far) no support for dependent types or higher-order abstract
syntax (which would be easy to add, but make interpreters much more difficult
to write)
Here is an example of a GF grammar, consisting of three modules,
as translated to GFCC. The representations are aligned, with the exceptions
due to the alphabetical sorting of GFCC grammars.
```
grammar Ex(Eng,Swe);
abstract Ex = { abstract {
cat
S ; NP ; VP ;
fun
Pred : NP -> VP -> S ; Pred : NP,VP -> S = (Pred);
She, They : NP ; She : -> NP = (She);
Sleep : VP ; Sleep : -> VP = (Sleep);
They : -> NP = (They);
} } ;
concrete Eng of Ex = { concrete Eng {
lincat
S = {s : Str} ;
NP = {s : Str ; n : Num} ;
VP = {s : Num => Str} ;
param
Num = Sg | Pl ;
lin
Pred np vp = { Pred = [(($0!1),(($1!0)!($0!0)))];
s = np.s ++ vp.s ! np.n} ;
She = {s = "she" ; n = Sg} ; She = [0, "she"];
They = {s = "they" ; n = Pl} ;
Sleep = {s = table { Sleep = [("sleep" + ["s",""])];
Sg => "sleeps" ;
Pl => "sleep" They = [1, "they"];
} } ;
} ;
}
concrete Swe of Ex = { concrete Swe {
lincat
S = {s : Str} ;
NP = {s : Str} ;
VP = {s : Str} ;
param
Num = Sg | Pl ;
lin
Pred np vp = { Pred = [(($0!0),($1!0))];
s = np.s ++ vp.s} ;
She = {s = "hon"} ; She = ["hon"];
They = {s = "de"} ; They = ["de"];
Sleep = {s = "sover"} ; Sleep = ["sover"];
} } ;
```
==The syntax of GFCC files==
===Top level===
A grammar has a header telling the name of the abstract syntax
(often specifying an application domain), and the names of
the concrete languages. The abstract syntax and the concrete
syntaxes themselves follow.
```
Grammar ::= Header ";" Abstract ";" [Concrete] ;
Header ::= "grammar" CId "(" [CId] ")" ;
Abstract ::= "abstract" "{" [AbsDef] "}" ;
Concrete ::= "concrete" CId "{" [CncDef] "}" ;
```
Abstract syntax judgements give typings and semantic definitions.
Concrete syntax judgements give linearizations.
```
AbsDef ::= CId ":" Type "=" Exp ;
CncDef ::= CId "=" Term ;
```
Also flags are possible, local to each "module" (i.e. abstract and concretes).
```
AbsDef ::= "%" CId "=" String ;
CncDef ::= "%" CId "=" String ;
```
For the run-time system, the reference implementation in Haskell
uses a structure that gives efficient look-up:
```
data GFCC = GFCC {
absname :: CId ,
cncnames :: [CId] ,
abstract :: Abstr ,
concretes :: Map CId Concr
}
data Abstr = Abstr {
funs :: Map CId Type, -- find the type of a fun
cats :: Map CId [CId] -- find the funs giving a cat
}
type Concr = Map CId Term
```
===Abstract syntax===
Types are first-order function types built from
category symbols. Syntax trees (``Exp``) are
rose trees with the head (``Atom``) either a function
constant, a metavariable, or a string, integer, or float
literal.
```
Type ::= [CId] "->" CId ;
Exp ::= "(" Atom [Exp] ")" ;
Atom ::= CId ; -- function constant
Atom ::= "?" ; -- metavariable
Atom ::= String ; -- string literal
Atom ::= Integer ; -- integer literal
Atom ::= Double ; -- float literal
```
===Concrete syntax===
Linearization terms (``Term``) are built as follows.
Constructor names are shown to make the later code
examples readable.
```
R. Term ::= "[" [Term] "]" ; -- array
P. Term ::= "(" Term "!" Term ")" ; -- access to indexed field
S. Term ::= "(" [Term] ")" ; -- sequence with ++
K. Term ::= Tokn ; -- token
V. Term ::= "$" Integer ; -- argument
C. Term ::= Integer ; -- array index
FV. Term ::= "[|" [Term] "|]" ; -- free variation
TM. Term ::= "?" ; -- linearization of metavariable
```
Tokens are strings or (maybe obsolescent) prefix-dependent
variant lists.
```
KS. Tokn ::= String ;
KP. Tokn ::= "[" "pre" [String] "[" [Variant] "]" "]" ;
Var. Variant ::= [String] "/" [String] ;
```
Three special forms of terms are introduced by the compiler
as optimizations. They can in principle be eliminated, but
their presence makes grammars much more compact. Their semantics
will be explained in a later section.
```
F. Term ::= CId ; -- global constant
W. Term ::= "(" String "+" Term ")" ; -- prefix + suffix table
RP. Term ::= "(" Term "@" Term ")"; -- record parameter alias
```
Identifiers are like ``Ident`` in GF and GFC, except that
the compiler produces constants prefixed with ``_`` in
the common subterm elimination optimization.
```
token CId (('_' | letter) (letter | digit | '\'' | '_')*) ;
```
==The semantics of concrete syntax terms==
===Linearization and realization===
The linearization algorithm is essentially the same as in
GFC: a tree is linearized by evaluating its linearization term
in the environment of the linearizations of the subtrees.
Literal atoms are linearized in the obvious way.
The function also needs to know the language (i.e. concrete syntax)
in which linearization is performed.
```
linExp :: GFCC -> CId -> Exp -> Term
linExp mcfg lang tree@(Tr at trees) = case at of
AC fun -> comp (Prelude.map lin trees) $ look fun
AS s -> R [kks (show s)] -- quoted
AI i -> R [kks (show i)]
AF d -> R [kks (show d)]
AM -> TM
where
lin = linExp mcfg lang
comp = compute mcfg lang
look = lookLin mcfg lang
```
The result of linearization is usually a record, which is realized as
a string using the following algorithm.
```
realize :: Term -> String
realize trm = case trm of
R (t:_) -> realize t
S ss -> unwords $ Prelude.map realize ss
K (KS s) -> s
K (KP s _) -> unwords s ---- prefix choice TODO
W s t -> s ++ realize t
FV (t:_) -> realize t
TM -> "?"
```
Since the order of record fields is not necessarily
the same as in GF source,
this realization does not work securely for
categories whose lincats more than one field.
===Term evaluation===
Evaluation follows call-by-value order, with two environments
needed:
- the grammar (a concrete syntax) to give the global constants
- an array of terms to give the subtree linearizations
The code is presented in one-level pattern matching, to
enable reimplementations in languages that do not permit
deep patterns (such as Java and C++).
```
compute :: GFCC -> CId -> [Term] -> Term -> Term
compute mcfg lang args = comp where
comp trm = case trm of
P r p -> proj (comp r) (comp p)
RP i t -> RP (comp i) (comp t)
W s t -> W s (comp t)
R ts -> R $ Prelude.map comp ts
V i -> idx args (fromInteger i) -- already computed
F c -> comp $ look c -- not computed (if contains V)
FV ts -> FV $ Prelude.map comp ts
S ts -> S $ Prelude.filter (/= S []) $ Prelude.map comp ts
_ -> trm
look = lookLin mcfg lang
idx xs i = xs !! i
proj r p = case (r,p) of
(_, FV ts) -> FV $ Prelude.map (proj r) ts
(W s t, _) -> kks (s ++ getString (proj t p))
_ -> comp $ getField r (getIndex p)
getString t = case t of
K (KS s) -> s
_ -> trace ("ERROR in grammar compiler: string from "++ show t) "ERR"
getIndex t = case t of
C i -> fromInteger i
RP p _ -> getIndex p
TM -> 0 -- default value for parameter
_ -> trace ("ERROR in grammar compiler: index from " ++ show t) 0
getField t i = case t of
R rs -> idx rs i
RP _ r -> getField r i
TM -> TM
_ -> trace ("ERROR in grammar compiler: field from " ++ show t) t
```
===The special term constructors===
The three forms introduced by the compiler may a need special
explanation.
Global constants
```
Term ::= CId ;
```
are shorthands for complex terms. They are produced by the
compiler by (iterated) common subexpression elimination.
They are often more powerful than hand-devised code sharing in the source
code. They could be computed off-line by replacing each identifier by
its definition.
Prefix-suffix tables
```
Term ::= "(" String "+" Term ")" ;
```
represent tables of word forms divided to the longest common prefix
and its array of suffixes. In the example grammar above, we have
```
Sleep = [("sleep" + ["s",""])]
```
which in fact is equal to the array of full forms
```
["sleeps", "sleep"]
```
The power of this construction comes from the fact that suffix sets
tend to be repeated in a language, and can therefore be collected
by common subexpression elimination. It is this technique that
explains the used syntax rather than the more accurate
```
"(" String "+" [String] ")"
```
since we want the suffix part to be a ``Term`` for the optimization to
take effect.
The most curious construct of GFCC is the parameter array alias,
```
Term ::= "(" Term "@" Term ")";
```
This form is used as the value of parameter records, such as the type
```
{n : Number ; p : Person}
```
The problem with parameter records is their double role.
They can be used like parameter values, as indices in selection,
```
VP.s ! {n = Sg ; p = P3}
```
but also as records, from which parameters can be projected:
```
{n = Sg ; p = P3}.n
```
Whichever use is selected as primary, a prohibitively complex
case expression must be generated at compilation to GFCC to get the
other use. The adopted
solution is to generate a pair containing both a parameter value index
and an array of indices of record fields. For instance, if we have
```
param Number = Sg | Pl ; Person = P1 | P2 | P3 ;
```
we get the encoding
```
{n = Sg ; p = P3} ---> (2 @ [0,2])
```
The GFCC computation rules are essentially
```
(t ! (i @ _)) = (t ! i)
((_ @ r) ! j) =(r ! j)
```
==Compiling to GFCC==
Compilation to GFCC is performed by the GF grammar compiler, and
GFCC interpreters need not know what it does. For grammar writers,
however, it might be interesting to know what happens to the grammars
in the process.
The compilation phases are the following
+ translate GF source to GFC, as always in GF
+ undo GFC back-end optimizations
+ perform the ``values`` optimization to normalize tables
+ create a symbol table mapping the GFC parameter and record types to
fixed-size arrays, and parameter values and record labels to integers
+ traverse the linearization rules replacing parameters and labels by integers
+ reorganize the created GFC grammar so that it has just one abstract syntax
and one concrete syntax per language
+ apply UTF8 encoding to the grammar, if not yet applied (this is told by the
``coding`` flag)
+ translate the GFC syntax tree to a GFCC syntax tree, using a simple
compositional mapping
+ perform the word-suffix optimization on GFCC linearization terms
+ perform subexpression elimination on each concrete syntax module
+ print out the GFCC code
Notice that a major part of the compilation is done within GFC, so that
GFC-related tasks (such as parser generation) could be performed by
using the old algorithms.
===Problems in GFCC compilation===
Two major problems had to be solved in compiling GFC to GFCC:
- consistent order of tables and records, to permit the array translation
- run-time variables in complex parameter values.
The current implementation is still experimental and may fail
to generate correct code. Any errors remaining are likely to be
related to the two problems just mentioned.
The order problem is solved in different ways for tables and records.
For tables, the ``values`` optimization of GFC already manages to
maintain a canonical order. But this order can be destroyed by the
``share`` optimization. To make sure that GFCC compilation works properly,
it is safest to recompile the GF grammar by using the ``values``
optimization flag.
Records can be canonically ordered by sorting them by labels.
In fact, this was done in connection of the GFCC work as a part
of the GFC generation, to guarantee consistency. This means that
e.g. the ``s`` field will in general no longer appear as the first
field, even if it does so in the GF source code. But relying on the
order of fields in a labelled record would be misplaced anyway.
The canonical form of records is further complicated by lock fields,
i.e. dummy fields of form ``lock_C = <>``, which are added to grammar
libraries to force intensionality of linearization types. The problem
is that the absence of a lock field only generates a warning, not
an error. Therefore a GFC grammar can contain objects of the same
type with and without a lock field. This problem was solved in GFCC
generation by just removing all lock fields (defined as fields whose
type is the empty record type). This has the further advantage of
(slightly) reducing the grammar size. More importantly, it is safe
to remove lock fields, because they are never used in computation,
and because intensional types are only needed in grammars reused
as libraries, not in grammars used at runtime.
While the order problem is rather bureaucratic in nature, run-time
variables are an interesting problem. They arise in the presence
of complex parameter values, created by argument-taking constructors
and parameter records. To give an example, consider the GF parameter
type system
```
Number = Sg | Pl ;
Person = P1 | P2 | P3 ;
Agr = Ag Number Person ;
```
The values can be translated to integers in the expected way,
```
Sg = 0, Pl = 1
P1 = 0, P2 = 1, P3 = 2
Ag Sg P1 = 0, Ag Sg P2 = 1, Ag Sg P3 = 2,
Ag Pl P1 = 3, Ag Pl P2 = 4, Ag Pl P3 = 5
```
However, an argument of ``Agr`` can be a run-time variable, as in
```
Ag np.n P3
```
This expression must first be translated to a case expression,
```
case np.n of {
0 => 2 ;
1 => 5
}
```
which can then be translated to the GFCC term
```
([2,5] ! ($0 ! $1))
```
assuming that the variable ``np`` is the first argument and that its
``Number`` field is the second in the record.
This transformation of course has to be performed recursively, since
there can be several run-time variables in a parameter value:
```
Ag np.n np.p
```
A similar transformation would be possible to deal with the double
role of parameter records discussed above. Thus the type
```
RNP = {n : Number ; p : Person}
```
could be uniformly translated into the set ``{0,1,2,3,4,5}``
as ``Agr`` above. Selections would be simple instances of indexing.
But any projection from the record should be translated into
a case expression,
```
rnp.n ===>
case rnp of {
0 => 0 ;
1 => 0 ;
2 => 0 ;
3 => 1 ;
4 => 1 ;
5 => 1
}
```
To avoid the code bloat resulting from this, we chose the alias representation
which is easy enough to deal with in interpreters.
===The representation of linearization types===
Linearization types (``lincat``) are not needed when generating with
GFCC, but they have been added to enable parser generation directly from
GFCC. The linearization type definitions are shown as a part of the
concrete syntax, by using terms to represent types. Here is the table
showing how different linearization types are encoded.
```
P* = size(P) -- parameter type
{_ : I ; __ : R}* = (I* @ R*) -- record of parameters
{r1 : T1 ; ... ; rn : Tn}* = [T1*,...,Tn*] -- other record
(P => T)* = [T* ,...,T*] -- size(P) times
Str* = ()
```
The category symbols are prefixed with two underscores (``__``).
For example, the linearization type ``present/CatEng.NP`` is
translated as follows:
```
NP = {
a : { -- 6 = 2*3 values
n : {ParamX.Number} ; -- 2 values
p : {ParamX.Person} -- 3 values
} ;
s : {ResEng.Case} => Str -- 3 values
}
__NP = [(6@[2,3]),[(),(),()]]
```
===Running the compiler and the GFCC interpreter===
GFCC generation is a part of the
[developers' version http://www.cs.chalmers.se/Cs/Research/Language-technology/darcs/GF/doc/darcs.html]
of GF since September 2006. To invoke the compiler, the flag
``-printer=gfcc`` to the command
``pm = print_multi`` is used. It is wise to recompile the grammar from
source, since previously compiled libraries may not obey the canonical
order of records. To ``strip`` the grammar before
GFCC translation removes unnecessary interface references.
Here is an example, performed in
[example/bronzeage ../../../../../examples/bronzeage].
```
i -src -path=.:prelude:resource-1.0/* -optimize=all_subs BronzeageEng.gf
i -src -path=.:prelude:resource-1.0/* -optimize=all_subs BronzeageGer.gf
strip
pm -printer=gfcc | wf bronze.gfcc
```
==The reference interpreter==
The reference interpreter written in Haskell consists of the following files:
```
-- source file for BNFC
GFCC.cf -- labelled BNF grammar of gfcc
-- files generated by BNFC
AbsGFCC.hs -- abstrac syntax of gfcc
ErrM.hs -- error monad used internally
LexGFCC.hs -- lexer of gfcc files
ParGFCC.hs -- parser of gfcc files and syntax trees
PrintGFCC.hs -- printer of gfcc files and syntax trees
-- hand-written files
DataGFCC.hs -- post-parser grammar creation, linearization and evaluation
GenGFCC.hs -- random and exhaustive generation, generate-and-test parsing
RunGFCC.hs -- main function - a simple command interpreter
```
It is included in the
[developers' version http://www.cs.chalmers.se/Cs/Research/Language-technology/darcs/GF/doc/darcs.html]
of GF, in the subdirectory [``GF/src/GF/Canon/GFCC`` ../].
To compile the interpreter, type
```
make gfcc
```
in ``GF/src``. To run it, type
```
./gfcc <GFCC-file>
```
The available commands are
- ``gr <Cat> <Int>``: generate a number of random trees in category.
and show their linearizations in all languages
- ``grt <Cat> <Int>``: generate a number of random trees in category.
and show the trees and their linearizations in all languages
- ``gt <Cat> <Int>``: generate a number of trees in category from smallest,
and show their linearizations in all languages
- ``gtt <Cat> <Int>``: generate a number of trees in category from smallest,
and show the trees and their linearizations in all languages
- ``p <Int> <Cat> <String>``: "parse", i.e. generate trees until match or
until the given number have been generated
- ``<Tree>``: linearize tree in all languages, also showing full records
- ``quit``: terminate the system cleanly
==Interpreter in C++==
A base-line interpreter in C++ has been started.
Its main functionality is random generation of trees and linearization of them.
Here are some results from running the different interpreters, compared
to running the same grammar in GF, saved in ``.gfcm`` format.
The grammar contains the English, German, and Norwegian
versions of Bronzeage. The experiment was carried out on
Ubuntu Linux laptop with 1.5 GHz Intel centrino processor.
|| | GF | gfcc(hs) | gfcc++ |
| program size | 7249k | 803k | 113k
| grammar size | 336k | 119k | 119k
| read grammar | 1150ms | 510ms | 100ms
| generate 222 | 9500ms | 450ms | 800ms
| memory | 21M | 10M | 20M
To summarize:
- going from GF to gfcc is a major win in both code size and efficiency
- going from Haskell to C++ interpreter is not a win yet, because of a space
leak in the C++ version
==Some things to do==
Interpreter in Java.
Parsing via MCFG
- the FCFG format can possibly be simplified
- parser grammars should be saved in files to make interpreters easier
Hand-written parsers for GFCC grammars to reduce code size
(and efficiency?) of interpreters.
Binary format and/or file compression of GFCC output.
Syntax editor based on GFCC.
Rewriting of resource libraries in order to exploit the
word-suffix sharing better (depth-one tables, as in FM).