From c316cc894443e526b4986cf5f778a4f544bae6a8 Mon Sep 17 00:00:00 2001
From: "Sven M. Hallberg" <pesco@khjk.org>
Date: Sat, 8 Feb 2020 20:10:49 +0100
Subject: [PATCH] let 'obj' no longer consume leading whitespace
---
pdf.c | 32 ++++++++++++++++++--------------
1 file changed, 18 insertions(+), 14 deletions(-)
diff --git a/pdf.c b/pdf.c
index 2ceff3c..208cd10 100644
--- a/pdf.c
+++ b/pdf.c
@@ -509,13 +509,14 @@ init_parser(struct Env *aux)
/* whitespace */
H_RULE(comment, SEQ(percent, line));
- H_RULE(ws, IGN(h_many(CHX(wchar, comment))));
- H_RULE(lws, IGN(h_many(lwchar)));
+ H_RULE(wel, IGN(CHX(wchar, comment));
+ H_RULE(ws, IGN(h_many(wel))));
+ H_RULE(lws, IGN(h_many(IGN(lwchar))));
+ /* tokenization */
+#define MANY_WS(X) h_many(CHX(wel, X))
#define TOK(X) h_middle(ws, X, h_not(rchar))
#define KW(S) IGN(TOK(LIT(S)))
-#define TOKD(X) h_right(ws, X) /* for tokens that end on delimiters */
-// XXX this allows, for instance, "<<<<" to be parsed as "<< <<". ok?
/* misc */
H_RULE(nl, IGN(h_right(lws, eol)));
@@ -524,6 +525,7 @@ init_parser(struct Env *aux)
H_RULE(empty, SEQ(epsilon));
H_ARULE(nat, TOK(h_many1(digit)));
H_ARULE(pnat, TOK(SEQ(pdigit, h_many(digit))));
+ //H_RULE(npair, SEQ(pnat, wel,ws, nat));
#define OPT(X) CHX(X, epsilon)
@@ -562,9 +564,8 @@ init_parser(struct Env *aux)
/* NB: lone backslashes and escaped newlines are ignored */
H_ARULE(schars, h_many(CHX(schar, snest, sesc, eol)));
H_RULE(snest_, SEQ(lparen, schars, rparen));
- H_RULE(litstr, h_middle(TOKD(lparen), schars, rparen));
- H_RULE(hexchr, h_right(ws, hdigit));
- H_RULE(hexstr, h_middle(TOKD(langle), h_many(hexchr), TOKD(rangle)));
+ H_RULE(litstr, h_middle(lparen, schars, rparen));
+ H_RULE(hexstr, h_middle(langle, MANY_WS(hdigit), rangle));
H_ARULE(string, CHX(litstr, hexstr));
h_bind_indirect(snest, snest_);
@@ -572,13 +573,16 @@ init_parser(struct Env *aux)
H_RULE(dopen, LIT("<<"));
H_RULE(dclose, LIT(">>"));
H_RULE(obj, h_indirect());
- H_RULE(k_v, SEQ(name, obj));
- H_ARULE(dict, h_middle(TOKD(dopen), h_many(k_v), TOKD(dclose)));
- H_RULE(array, h_middle(TOKD(lbrack), h_many(obj), TOKD(rbrack)));
+ H_RULE(k_v, SEQ(name, ws, obj));
+ //H_RULE(k_v, CHX(SEQ(name, wel,ws, obj),
+ // SEQ(name, dobj)));
+ H_ARULE(dict, h_middle(dopen, MANY_WS(k_v), dclose));
+ // XXX this allows, for instance, "<<<<" to be parsed as "<< <<". ok?
+ H_RULE(array, h_middle(lbrack, MANY_WS(obj), rbrack));
// XXX validate: dict keys must be unique
/* streams */
- H_RULE(stmbeg, SEQ(dict, KW("stream"), OPT(cr), lf));
+ H_RULE(stmbeg, SEQ(dict, ws, LIT("stream"), OPT(cr), lf));
H_RULE(stmend, SEQ(OPT(eol), LIT("endstream")));
H_RULE(stream, h_left(h_bind(stmbeg, kstream, aux), stmend));
// XXX is whitespace allowed between the eol and "endstream"?
@@ -597,7 +601,7 @@ init_parser(struct Env *aux)
/* body */
H_RULE(indobj, CHX(stream, obj));
- H_RULE(objdef, SEQ(pnat, nat, KW("obj"), indobj, KW("endobj")));
+ H_RULE(objdef, SEQ(pnat, nat, KW("obj"), ws, indobj, KW("endobj")));
H_RULE(body, h_many(objdef));
/* for object streams */
@@ -617,7 +621,7 @@ init_parser(struct Env *aux)
/* cross-reference streams */
H_RULE(xstream, h_bind(stmbeg, kxstream, aux));
- H_AVRULE(xrstm, SEQ(pnat, nat, KW("obj"), xstream));
+ H_AVRULE(xrstm, SEQ(pnat, nat, KW("obj"), ws, xstream));
// XXX skip however much we consumed and check for "endstream endobj"?
/* trailer */
@@ -626,7 +630,7 @@ init_parser(struct Env *aux)
LIT("%%EOF"), CHX(nl, end)));
// XXX should lws be allowed before EOF marker?
// NB: lws before xref offset is allowed, cf. p.48 (example 4)
- H_RULE(xr_td, SEQ(xrefs, KW("trailer"), dict));
+ H_RULE(xr_td, SEQ(xrefs, KW("trailer"), ws, dict));
H_RULE(tail, SEQ(body, h_optional(xr_td), startxr));
H_RULE(pdf, SEQ(header, h_many1(tail), end));
--
GitLab