\end{verbatim}
\item
and finally: (ToDo: fill in)
The right way to think about the ``after-match function'' is that it
is an embryonic @CoreExpr@ with a ``hole'' at the end for the
final ``else expression''.
\end{itemize}
There is a type synonym, @EquationInfo@, defined in module @DsUtils@.
An experiment with re-ordering this information about equations (in
particular, having the patterns available in column-major order)
showed no benefit.
\item
A default expression---what to evaluate if the overall pattern-match
fails. This expression will (almost?) always be
a measly expression @Var@, unless we know it will only be used once
(as we do in @glue_success_exprs@).
Leaving out this third argument to @match@ (and slamming in lots of
@Var "fail"@s) is a positively {\em bad} idea, because it makes it
impossible to share the default expressions. (Also, it stands no
chance of working in our post-upheaval world of @Locals@.)
\end{enumerate}
Note: @match@ is often called via @matchWrapper@ (end of this module),
a function that does much of the house-keeping that goes with a call
to @match@.
It is also worth mentioning the {\em typical} way a block of equations
is desugared with @match@. At each stage, it is the first column of
patterns that is examined. The steps carried out are roughly:
\begin{enumerate}
\item
Tidy the patterns in column~1 with @tidyEqnInfo@ (this may add
bindings to the second component of the equation-info):
\begin{itemize}
\item
Remove the `as' patterns from column~1.
\item
Make all constructor patterns in column~1 into @ConPats@, notably
@ListPats@ and @TuplePats@.
\item
Handle any irrefutable (or ``twiddle'') @LazyPats@.
\end{itemize}
\item
Now {\em unmix} the equations into {\em blocks} [w\/ local function
@unmix_eqns@], in which the equations in a block all have variable
patterns in column~1, or they all have constructor patterns in ...
(see ``the mixture rule'' in SLPJ).
\item
Call @matchEqnBlock@ on each block of equations; it will do the
appropriate thing for each kind of column-1 pattern, usually ending up
in a recursive call to @match@.
\end{enumerate}
We are a little more paranoid about the ``empty rule'' (SLPJ, p.~87)
than the Wadler-chapter code for @match@ (p.~93, first @match@ clause).
And gluing the ``success expressions'' together isn't quite so pretty.
This (more interesting) clause of @match@ uses @tidy_and_unmix_eqns@
(a)~to get `as'- and `twiddle'-patterns out of the way (tidying), and
(b)~to do ``the mixture rule'' (SLPJ, p.~88) [which really {\em
un}mixes the equations], producing a list of equation-info
blocks, each block having as its first column of patterns either all
constructors, or all variables (or similar beasts), etc.
@match_unmixed_eqn_blks@ simply takes the place of the @foldr@ in the
Wadler-chapter @match@ (p.~93, last clause), and @match_unmixed_blk@
corresponds roughly to @matchVarCon@.
\begin{code}
match :: [Id]
-> Type
-> [EquationInfo]
-> DsM MatchResult
match [] ty eqns
= ASSERT2( not (null eqns), ppr ty )
return (foldr1 combineMatchResults match_results)
where
match_results = [ ASSERT( null (eqn_pats eqn) )
eqn_rhs eqn
| eqn <- eqns ]
match vars@(v:_) ty eqns
= ASSERT( not (null eqns ) )
do {
(aux_binds, tidy_eqns) <- mapAndUnzipM (tidyEqnInfo v) eqns
; let grouped = groupEquations tidy_eqns
; ifDOptM Opt_D_dump_view_pattern_commoning (debug grouped)
; match_results <- mapM match_group grouped
; return (adjustMatchResult (foldr1 (.) aux_binds) $
foldr1 combineMatchResults match_results) }
where
dropGroup :: [(PatGroup,EquationInfo)] -> [EquationInfo]
dropGroup = map snd
match_group :: [(PatGroup,EquationInfo)] -> DsM MatchResult
match_group [] = panic "match_group"
match_group eqns@((group,_) : _)
= case group of
PgCon _ -> matchConFamily vars ty (subGroup [(c,e) | (PgCon c, e) <- eqns])
PgLit _ -> matchLiterals vars ty (subGroup [(l,e) | (PgLit l, e) <- eqns])
PgAny -> matchVariables vars ty (dropGroup eqns)
PgN _ -> matchNPats vars ty (dropGroup eqns)
PgNpK _ -> matchNPlusKPats vars ty (dropGroup eqns)
PgBang -> matchBangs vars ty (dropGroup eqns)
PgCo _ -> matchCoercion vars ty (dropGroup eqns)
PgView _ _ -> matchView vars ty (dropGroup eqns)
debug eqns =
let gs = map (\group -> foldr (\ (p,_) -> \acc ->
case p of PgView e _ -> e:acc
_ -> acc) [] group) eqns
maybeWarn [] = return ()
maybeWarn l = warnDs (vcat l)
in
maybeWarn $ (map (\g -> text "Putting these view expressions into the same case:" <+> (ppr g))
(filter (not . null) gs))
matchVariables :: [Id] -> Type -> [EquationInfo] -> DsM MatchResult
matchVariables (_:vars) ty eqns = match vars ty (shiftEqns eqns)
matchVariables [] _ _ = panic "matchVariables"
matchBangs :: [Id] -> Type -> [EquationInfo] -> DsM MatchResult
matchBangs (var:vars) ty eqns
= do { match_result <- match (var:vars) ty $
map (decomposeFirstPat getBangPat) eqns
; return (mkEvalMatchResult var ty match_result) }
matchBangs [] _ _ = panic "matchBangs"
matchCoercion :: [Id] -> Type -> [EquationInfo] -> DsM MatchResult
matchCoercion (var:vars) ty (eqns@(eqn1:_))
= do { let CoPat co pat _ = firstPat eqn1
; var' <- newUniqueId var (hsPatType pat)
; match_result <- match (var':vars) ty $
map (decomposeFirstPat getCoPat) eqns
; co' <- dsHsWrapper co
; let rhs' = co' (Var var)
; return (mkCoLetMatchResult (NonRec var' rhs') match_result) }
matchCoercion _ _ _ = panic "matchCoercion"
matchView :: [Id] -> Type -> [EquationInfo] -> DsM MatchResult
matchView (var:vars) ty (eqns@(eqn1:_))
= do {
let ViewPat viewExpr (L _ pat) _ = firstPat eqn1
; var' <- newUniqueId var (hsPatType pat)
; match_result <- match (var':vars) ty $
map (decomposeFirstPat getViewPat) eqns
; viewExpr' <- dsLExpr viewExpr
; return (mkViewMatchResult var' viewExpr' var match_result) }
matchView _ _ _ = panic "matchView"
decomposeFirstPat :: (Pat Id -> Pat Id) -> EquationInfo -> EquationInfo
decomposeFirstPat extractpat (eqn@(EqnInfo { eqn_pats = pat : pats }))
= eqn { eqn_pats = extractpat pat : pats}
decomposeFirstPat _ _ = panic "decomposeFirstPat"
getCoPat, getBangPat, getViewPat :: Pat Id -> Pat Id
getCoPat (CoPat _ pat _) = pat
getCoPat _ = panic "getCoPat"
getBangPat (BangPat pat ) = unLoc pat
getBangPat _ = panic "getBangPat"
getViewPat (ViewPat _ pat _) = unLoc pat
getViewPat _ = panic "getBangPat"
\end{code}
%************************************************************************
%* *
Tidying patterns
%* *
%************************************************************************
Tidy up the leftmost pattern in an @EquationInfo@, given the variable @v@
which will be scrutinised. This means:
\begin{itemize}
\item
Replace variable patterns @x@ (@x /= v@) with the pattern @_@,
together with the binding @x = v@.
\item
Replace the `as' pattern @x@@p@ with the pattern p and a binding @x = v@.
\item
Removing lazy (irrefutable) patterns (you don't want to know...).
\item
Converting explicit tuple-, list-, and parallel-array-pats into ordinary
@ConPats@.
\item
Convert the literal pat "" to [].
\end{itemize}
The result of this tidying is that the column of patterns will include
{\em only}:
\begin{description}
\item[@WildPats@:]
The @VarPat@ information isn't needed any more after this.
\item[@ConPats@:]
@ListPats@, @TuplePats@, etc., are all converted into @ConPats@.
\item[@LitPats@ and @NPats@:]
@LitPats@/@NPats@ of ``known friendly types'' (Int, Char,
Float, Double, at least) are converted to unboxed form; e.g.,
\tr{(NPat (HsInt i) _ _)} is converted to:
\begin{verbatim}
(ConPat I# _ _ [LitPat (HsIntPrim i)])
\end{verbatim}
\end{description}
\begin{code}
tidyEqnInfo :: Id -> EquationInfo
-> DsM (DsWrapper, EquationInfo)
tidyEqnInfo _ (EqnInfo { eqn_pats = [] })
= panic "tidyEqnInfo"
tidyEqnInfo v eqn@(EqnInfo { eqn_pats = pat : pats })
= do { (wrap, pat') <- tidy1 v pat
; return (wrap, eqn { eqn_pats = do pat' : pats }) }
tidy1 :: Id
-> Pat Id
-> DsM (DsWrapper,
Pat Id)
tidy1 v (ParPat pat) = tidy1 v (unLoc pat)
tidy1 v (SigPatOut pat _) = tidy1 v (unLoc pat)
tidy1 _ (WildPat ty) = return (idDsWrapper, WildPat ty)
tidy1 v (VarPat var)
= return (wrapBind var v, WildPat (idType var))
tidy1 v (VarPatOut var binds)
= do { ds_ev_binds <- dsTcEvBinds binds
; return (wrapBind var v . wrapDsEvBinds ds_ev_binds,
WildPat (idType var)) }
tidy1 v (AsPat (L _ var) pat)
= do { (wrap, pat') <- tidy1 v (unLoc pat)
; return (wrapBind var v . wrap, pat') }
tidy1 v (LazyPat pat)
= do { sel_prs <- mkSelectorBinds pat (Var v)
; let sel_binds = [NonRec b rhs | (b,rhs) <- sel_prs]
; return (mkCoreLets sel_binds, WildPat (idType v)) }
tidy1 _ (ListPat pats ty)
= return (idDsWrapper, unLoc list_ConPat)
where
list_ty = mkListTy ty
list_ConPat = foldr (\ x y -> mkPrefixConPat consDataCon [x, y] list_ty)
(mkNilPat list_ty)
pats
tidy1 _ (PArrPat pats ty)
= return (idDsWrapper, unLoc parrConPat)
where
arity = length pats
parrConPat = mkPrefixConPat (parrFakeCon arity) pats (mkPArrTy ty)
tidy1 _ (TuplePat pats boxity ty)
= return (idDsWrapper, unLoc tuple_ConPat)
where
arity = length pats
tuple_ConPat = mkPrefixConPat (tupleCon boxity arity) pats ty
tidy1 _ (LitPat lit)
= return (idDsWrapper, tidyLitPat lit)
tidy1 _ (NPat lit mb_neg eq)
= return (idDsWrapper, tidyNPat lit mb_neg eq)
tidy1 v (BangPat (L _ (LazyPat p))) = tidy1 v (BangPat p)
tidy1 v (BangPat (L _ (ParPat p))) = tidy1 v (BangPat p)
tidy1 _ p@(BangPat (L _(VarPat _))) = return (idDsWrapper, p)
tidy1 _ p@(BangPat (L _(VarPatOut _ _))) = return (idDsWrapper, p)
tidy1 _ p@(BangPat (L _ (WildPat _))) = return (idDsWrapper, p)
tidy1 _ p@(BangPat (L _ (CoPat _ _ _))) = return (idDsWrapper, p)
tidy1 _ p@(BangPat (L _ (SigPatIn _ _))) = return (idDsWrapper, p)
tidy1 _ p@(BangPat (L _ (SigPatOut _ _))) = return (idDsWrapper, p)
tidy1 v (BangPat (L _ (AsPat (L _ var) pat)))
= do { (wrap, pat') <- tidy1 v (BangPat pat)
; return (wrapBind var v . wrap, pat') }
tidy1 v (BangPat (L _ p)) = tidy1 v p
tidy1 _ non_interesting_pat
= return (idDsWrapper, non_interesting_pat)
\end{code}
\noindent
{\bf Previous @matchTwiddled@ stuff:}
Now we get to the only interesting part; note: there are choices for
translation [from Simon's notes]; translation~1:
\begin{verbatim}
deTwiddle [s,t] e
\end{verbatim}
returns
\begin{verbatim}
[ w = e,
s = case w of [s,t] -> s
t = case w of [s,t] -> t
]
\end{verbatim}
Here \tr{w} is a fresh variable, and the \tr{w}-binding prevents multiple
evaluation of \tr{e}. An alternative translation (No.~2):
\begin{verbatim}
[ w = case e of [s,t] -> (s,t)
s = case w of (s,t) -> s
t = case w of (s,t) -> t
]
\end{verbatim}
%************************************************************************
%* *
\subsubsection[improved-unmixing]{UNIMPLEMENTED idea for improved unmixing}
%* *
%************************************************************************
We might be able to optimise unmixing when confronted by
only-one-constructor-possible, of which tuples are the most notable
examples. Consider:
\begin{verbatim}
f (a,b,c) ... = ...
f d ... (e:f) = ...
f (g,h,i) ... = ...
f j ... = ...
\end{verbatim}
This definition would normally be unmixed into four equation blocks,
one per equation. But it could be unmixed into just one equation
block, because if the one equation matches (on the first column),
the others certainly will.
You have to be careful, though; the example
\begin{verbatim}
f j ... = ...
-------------------
f (a,b,c) ... = ...
f d ... (e:f) = ...
f (g,h,i) ... = ...
\end{verbatim}
{\em must} be broken into two blocks at the line shown; otherwise, you
are forcing unnecessary evaluation. In any case, the top-left pattern
always gives the cue. You could then unmix blocks into groups of...
\begin{description}
\item[all variables:]
As it is now.
\item[constructors or variables (mixed):]
Need to make sure the right names get bound for the variable patterns.
\item[literals or variables (mixed):]
Presumably just a variant on the constructor case (as it is now).
\end{description}
%************************************************************************
%* *
%* matchWrapper: a convenient way to call @match@ *
%* *
%************************************************************************
\subsection[matchWrapper]{@matchWrapper@: a convenient interface to @match@}
Calls to @match@ often involve similar (non-trivial) work; that work
is collected here, in @matchWrapper@. This function takes as
arguments:
\begin{itemize}
\item
Typchecked @Matches@ (of a function definition, or a case or lambda
expression)---the main input;
\item
An error message to be inserted into any (runtime) pattern-matching
failure messages.
\end{itemize}
As results, @matchWrapper@ produces:
\begin{itemize}
\item
A list of variables (@Locals@) that the caller must ``promise'' to
bind to appropriate values; and
\item
a @CoreExpr@, the desugared output (main result).
\end{itemize}
The main actions of @matchWrapper@ include:
\begin{enumerate}
\item
Flatten the @[TypecheckedMatch]@ into a suitable list of
@EquationInfo@s.
\item
Create as many new variables as there are patterns in a pattern-list
(in any one of the @EquationInfo@s).
\item
Create a suitable ``if it fails'' expression---a call to @error@ using
the error-string input; the {\em type} of this fail value can be found
by examining one of the RHS expressions in one of the @EquationInfo@s.
\item
Call @match@ with all of this information!
\end{enumerate}
\begin{code}
matchWrapper :: HsMatchContext Name
-> MatchGroup Id
-> DsM ([Id], CoreExpr)
\end{code}
There is one small problem with the Lambda Patterns, when somebody
writes something similar to:
\begin{verbatim}
(\ (x:xs) -> ...)
\end{verbatim}
he/she don't want a warning about incomplete patterns, that is done with
the flag @opt_WarnSimplePatterns@.
This problem also appears in the:
\begin{itemize}
\item @do@ patterns, but if the @do@ can fail
it creates another equation if the match can fail
(see @DsExpr.doDo@ function)
\item @let@ patterns, are treated by @matchSimply@
List Comprension Patterns, are treated by @matchSimply@ also
\end{itemize}
We can't call @matchSimply@ with Lambda patterns,
due to the fact that lambda patterns can have more than
one pattern, and match simply only accepts one pattern.
JJQC 30-Nov-1997
\begin{code}
matchWrapper ctxt (MatchGroup matches match_ty)
= ASSERT( notNull matches )
do { eqns_info <- mapM mk_eqn_info matches
; new_vars <- selectMatchVars arg_pats
; result_expr <- matchEquations ctxt new_vars eqns_info rhs_ty
; return (new_vars, result_expr) }
where
arg_pats = map unLoc (hsLMatchPats (head matches))
n_pats = length arg_pats
(_, rhs_ty) = splitFunTysN n_pats match_ty
mk_eqn_info (L _ (Match pats _ grhss))
= do { let upats = map unLoc pats
; match_result <- dsGRHSs ctxt upats grhss rhs_ty
; return (EqnInfo { eqn_pats = upats, eqn_rhs = match_result}) }
matchEquations :: HsMatchContext Name
-> [Id] -> [EquationInfo] -> Type
-> DsM CoreExpr
matchEquations ctxt vars eqns_info rhs_ty
= do { locn <- getSrcSpanDs
; let ds_ctxt = DsMatchContext ctxt locn
error_doc = matchContextErrString ctxt
; match_result <- matchCheck ds_ctxt vars rhs_ty eqns_info
; fail_expr <- mkErrorAppDs pAT_ERROR_ID rhs_ty error_doc
; extractMatchResult match_result fail_expr }
\end{code}
%************************************************************************
%* *
\subsection[matchSimply]{@matchSimply@: match a single expression against a single pattern}
%* *
%************************************************************************
@mkSimpleMatch@ is a wrapper for @match@ which deals with the
situation where we want to match a single expression against a single
pattern. It returns an expression.
\begin{code}
matchSimply :: CoreExpr
-> HsMatchContext Name
-> LPat Id
-> CoreExpr
-> CoreExpr
-> DsM CoreExpr
matchSimply scrut hs_ctx pat result_expr fail_expr = do
let
match_result = cantFailMatchResult result_expr
rhs_ty = exprType fail_expr
match_result' <- matchSinglePat scrut hs_ctx pat rhs_ty match_result
extractMatchResult match_result' fail_expr
matchSinglePat :: CoreExpr -> HsMatchContext Name -> LPat Id
-> Type -> MatchResult -> DsM MatchResult
matchSinglePat (Var var) _ (L _ pat) ty match_result
= match [var] ty [EqnInfo { eqn_pats = [pat], eqn_rhs = match_result }]
matchSinglePat scrut hs_ctx pat ty match_result = do
var <- selectSimpleMatchVarL pat
match_result' <- matchSinglePat (Var var) hs_ctx pat ty match_result
return (adjustMatchResult (bindNonRec var scrut) match_result')
\end{code}
%************************************************************************
%* *
Pattern classification
%* *
%************************************************************************
\begin{code}
data PatGroup
= PgAny
| PgCon DataCon
| PgLit Literal
| PgN Literal
| PgNpK Literal
| PgBang
| PgCo Type
| PgView (LHsExpr Id)
Type
groupEquations :: [EquationInfo] -> [[(PatGroup, EquationInfo)]]
groupEquations eqns
= runs same_gp [(patGroup (firstPat eqn), eqn) | eqn <- eqns]
where
same_gp :: (PatGroup,EquationInfo) -> (PatGroup,EquationInfo) -> Bool
(pg1,_) `same_gp` (pg2,_) = pg1 `sameGroup` pg2
subGroup :: Ord a => [(a, EquationInfo)] -> [[EquationInfo]]
subGroup group
= map reverse $ Map.elems $ foldl accumulate Map.empty group
where
accumulate pg_map (pg, eqn)
= case Map.lookup pg pg_map of
Just eqns -> Map.insert pg (eqn:eqns) pg_map
Nothing -> Map.insert pg [eqn] pg_map
\end{code}
Note [Take care with pattern order]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In the subGroup function we must be very careful about pattern re-ordering,
Consider the patterns [ (True, Nothing), (False, x), (True, y) ]
Then in bringing together the patterns for True, we must not
swap the Nothing and y!
\begin{code}
sameGroup :: PatGroup -> PatGroup -> Bool
sameGroup PgAny PgAny = True
sameGroup PgBang PgBang = True
sameGroup (PgCon _) (PgCon _) = True
sameGroup (PgLit _) (PgLit _) = True
sameGroup (PgN l1) (PgN l2) = l1==l2
sameGroup (PgNpK l1) (PgNpK l2) = l1==l2
sameGroup (PgCo t1) (PgCo t2) = t1 `coreEqType` t2
sameGroup (PgView e1 t1) (PgView e2 t2) = viewLExprEq (e1,t1) (e2,t2)
sameGroup _ _ = False
viewLExprEq :: (LHsExpr Id,Type) -> (LHsExpr Id,Type) -> Bool
viewLExprEq (e1,_) (e2,_) = lexp e1 e2
where
lexp :: LHsExpr Id -> LHsExpr Id -> Bool
lexp e e' = exp (unLoc e) (unLoc e')
exp :: HsExpr Id -> HsExpr Id -> Bool
exp (HsPar (L _ e)) e' = exp e e'
exp e (HsPar (L _ e')) = exp e e'
exp (HsWrap h e) (HsWrap h' e') = wrap h h' && exp e e'
exp (HsVar i) (HsVar i') = i == i'
exp (HsIPVar i) (HsIPVar i') = i == i'
exp (HsOverLit l) (HsOverLit l') =
tcEqType (overLitType l) (overLitType l') && l == l'
exp (HsApp e1 e2) (HsApp e1' e2') = lexp e1 e1' && lexp e2 e2'
exp (OpApp l o _ ri) (OpApp l' o' _ ri') =
lexp l l' && lexp o o' && lexp ri ri'
exp (NegApp e n) (NegApp e' n') = lexp e e' && exp n n'
exp (SectionL e1 e2) (SectionL e1' e2') =
lexp e1 e1' && lexp e2 e2'
exp (SectionR e1 e2) (SectionR e1' e2') =
lexp e1 e1' && lexp e2 e2'
exp (ExplicitTuple es1 _) (ExplicitTuple es2 _) =
eq_list tup_arg es1 es2
exp (HsIf _ e e1 e2) (HsIf _ e' e1' e2') =
lexp e e' && lexp e1 e1' && lexp e2 e2'
exp _ _ = False
tup_arg (Present e1) (Present e2) = lexp e1 e2
tup_arg (Missing t1) (Missing t2) = tcEqType t1 t2
tup_arg _ _ = False
wrap :: HsWrapper -> HsWrapper -> Bool
wrap WpHole WpHole = True
wrap (WpCompose w1 w2) (WpCompose w1' w2') = wrap w1 w1' && wrap w2 w2'
wrap (WpCast c) (WpCast c') = tcEqType c c'
wrap (WpEvApp et1) (WpEvApp et2) = ev_term et1 et2
wrap (WpTyApp t) (WpTyApp t') = tcEqType t t'
wrap _ _ = False
ev_term :: EvTerm -> EvTerm -> Bool
ev_term (EvId a) (EvId b) = a==b
ev_term (EvCoercion a) (EvCoercion b) = tcEqType a b
ev_term _ _ = False
eq_list :: (a->a->Bool) -> [a] -> [a] -> Bool
eq_list _ [] [] = True
eq_list _ [] (_:_) = False
eq_list _ (_:_) [] = False
eq_list eq (x:xs) (y:ys) = eq x y && eq_list eq xs ys
patGroup :: Pat Id -> PatGroup
patGroup (WildPat {}) = PgAny
patGroup (BangPat {}) = PgBang
patGroup (ConPatOut { pat_con = dc }) = PgCon (unLoc dc)
patGroup (LitPat lit) = PgLit (hsLitKey lit)
patGroup (NPat olit mb_neg _) = PgN (hsOverLitKey olit (isJust mb_neg))
patGroup (NPlusKPat _ olit _ _) = PgNpK (hsOverLitKey olit False)
patGroup (CoPat _ p _) = PgCo (hsPatType p)
patGroup (ViewPat expr p _) = PgView expr (hsPatType (unLoc p))
patGroup pat = pprPanic "patGroup" (ppr pat)
\end{code}
Note [Grouping overloaded literal patterns]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
WATCH OUT! Consider
f (n+1) = ...
f (n+2) = ...
f (n+1) = ...
We can't group the first and third together, because the second may match
the same thing as the first. Same goes for *overloaded* literal patterns
f 1 True = ...
f 2 False = ...
f 1 False = ...
If the first arg matches '1' but the second does not match 'True', we
cannot jump to the third equation! Because the same argument might
match '2'!
Hence we don't regard 1 and 2, or (n+1) and (n+2), as part of the same group.