[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[VM] [COMMIT XEMACS] Pass character count from coding systems to buffer
|
From: |
Aidan Kehoe |
|
Subject: |
[VM] [COMMIT XEMACS] Pass character count from coding systems to buffer insertion code. |
|
Date: |
Thu, 16 Jan 2014 17:10:35 +0000 |
APPROVE COMMIT
SUPERSEDES address@hidden
This doesn’t include character_tell() implementations for lots of the coding
systems where said implementation is easy, e.g. in iso-8859-1. Nor does it
include it for chain coding systems, e.g. those for non-Unix line endings. I
haven’t profiled it to the level I did for the first patch, should do that
soon.
# HG changeset patch
# User Aidan Kehoe <address@hidden>
# Date 1389889672 0
# Node ID 65d65b52d608ca1f17365a96fc3cf710a3af625c
# Parent 4004c3266c09888a9935242a462beb3fb28e02a3
Pass character count from coding systems to buffer insertion code.
src/ChangeLog addition:
2014-01-16 Aidan Kehoe <address@hidden>
Pass character count information from the no-conversion and
unicode coding systems to the buffer insertion code, making
#'find-file on large buffers a little snappier (if
ERROR_CHECK_TEXT is not defined).
* file-coding.c:
* file-coding.c (coding_character_tell): New.
* file-coding.c (conversion_coding_stream_description): New.
* file-coding.c (no_conversion_convert):
Update characters_seen when decoding.
* file-coding.c (no_conversion_character_tell): New.
* file-coding.c (lstream_type_create_file_coding): Create the
no_conversion type with data.
* file-coding.c (coding_system_type_create):
Make the character_tell method available here.
* file-coding.h:
* file-coding.h (struct coding_system_methods):
Add a new character_tell() method, passing charcount information
from the coding systems to the buffer code, avoiding duplicate
bytecount-to-charcount work especially with large buffers.
* fileio.c (Finsert_file_contents_internal):
Update this to pass charcount information to
buffer_insert_string_1(), if that is available from the lstream code.
* insdel.c:
* insdel.c (buffer_insert_string_1):
Add a new CCLEN argument, giving the character count of the string
to insert. It can be -1 to indicate that te function should work
it out itself using bytecount_to_charcount(), as it used to.
* insdel.c (buffer_insert_raw_string_1):
* insdel.c (buffer_insert_lisp_string_1):
* insdel.c (buffer_insert_ascstring_1):
* insdel.c (buffer_insert_emacs_char_1):
* insdel.c (buffer_insert_from_buffer_1):
* insdel.c (buffer_replace_char):
Update these functions to use the new calling convention.
* insdel.h:
* insdel.h (buffer_insert_string):
Update this header to reflect the new buffer_insert_string_1()
argument.
* lstream.c (Lstream_character_tell): New.
Return the number of characters *read* and seen by the consumer so
far, taking into account the unget buffer, and buffered reading.
* lstream.c (Lstream_unread):
Update unget_character_count here as appropriate.
* lstream.c (Lstream_rewind):
Reset unget_character_count here too.
* lstream.h:
* lstream.h (struct lstream):
Provide the character_tell method, add a new field,
unget_character_count, giving the number of characters ever passed
to Lstream_unread().
Declare Lstream_character_tell().
Make Lstream_ungetc(), which happens to be unused, an inline
function rather than a macro, in the course of updating it to
modify unget_character_count.
* print.c (output_string):
Use the new argument to buffer_insert_string_1().
* tests.c:
* tests.c (Ftest_character_tell):
New test function.
* tests.c (syms_of_tests):
Make it available.
* unicode.c:
* unicode.c (struct unicode_coding_stream):
* unicode.c (unicode_character_tell):
New method.
* unicode.c (unicode_convert):
Update the character counter as appropriate.
* unicode.c (coding_system_type_create_unicode):
Make the character_tell method available.
diff -r 4004c3266c09 -r 65d65b52d608 src/ChangeLog
--- a/src/ChangeLog Sun Dec 22 10:36:33 2013 +0000
+++ b/src/ChangeLog Thu Jan 16 16:27:52 2014 +0000
@@ -1,3 +1,82 @@
+2014-01-16 Aidan Kehoe <address@hidden>
+
+ Pass character count information from the no-conversion and
+ unicode coding systems to the buffer insertion code, making
+ #'find-file on large buffers a little snappier (if
+ ERROR_CHECK_TEXT is not defined).
+
+ * file-coding.c:
+ * file-coding.c (coding_character_tell): New.
+ * file-coding.c (conversion_coding_stream_description): New.
+ * file-coding.c (no_conversion_convert):
+ Update characters_seen when decoding.
+ * file-coding.c (no_conversion_character_tell): New.
+ * file-coding.c (lstream_type_create_file_coding): Create the
+ no_conversion type with data.
+ * file-coding.c (coding_system_type_create):
+ Make the character_tell method available here.
+ * file-coding.h:
+ * file-coding.h (struct coding_system_methods):
+ Add a new character_tell() method, passing charcount information
+ from the coding systems to the buffer code, avoiding duplicate
+ bytecount-to-charcount work especially with large buffers.
+
+ * fileio.c (Finsert_file_contents_internal):
+ Update this to pass charcount information to
+ buffer_insert_string_1(), if that is available from the lstream code.
+
+ * insdel.c:
+ * insdel.c (buffer_insert_string_1):
+ Add a new CCLEN argument, giving the character count of the string
+ to insert. It can be -1 to indicate that te function should work
+ it out itself using bytecount_to_charcount(), as it used to.
+ * insdel.c (buffer_insert_raw_string_1):
+ * insdel.c (buffer_insert_lisp_string_1):
+ * insdel.c (buffer_insert_ascstring_1):
+ * insdel.c (buffer_insert_emacs_char_1):
+ * insdel.c (buffer_insert_from_buffer_1):
+ * insdel.c (buffer_replace_char):
+ Update these functions to use the new calling convention.
+ * insdel.h:
+ * insdel.h (buffer_insert_string):
+ Update this header to reflect the new buffer_insert_string_1()
+ argument.
+
+ * lstream.c (Lstream_character_tell): New.
+ Return the number of characters *read* and seen by the consumer so
+ far, taking into account the unget buffer, and buffered reading.
+
+ * lstream.c (Lstream_unread):
+ Update unget_character_count here as appropriate.
+ * lstream.c (Lstream_rewind):
+ Reset unget_character_count here too.
+
+ * lstream.h:
+ * lstream.h (struct lstream):
+ Provide the character_tell method, add a new field,
+ unget_character_count, giving the number of characters ever passed
+ to Lstream_unread().
+ Declare Lstream_character_tell().
+ Make Lstream_ungetc(), which happens to be unused, an inline
+ function rather than a macro, in the course of updating it to
+ modify unget_character_count.
+
+ * print.c (output_string):
+ Use the new argument to buffer_insert_string_1().
+ * tests.c:
+ * tests.c (Ftest_character_tell):
+ New test function.
+ * tests.c (syms_of_tests):
+ Make it available.
+ * unicode.c:
+ * unicode.c (struct unicode_coding_stream):
+ * unicode.c (unicode_character_tell):
+ New method.
+ * unicode.c (unicode_convert):
+ Update the character counter as appropriate.
+ * unicode.c (coding_system_type_create_unicode):
+ Make the character_tell method available.
+
2013-12-19 Aidan Kehoe <address@hidden>
* text.c:
diff -r 4004c3266c09 -r 65d65b52d608 src/file-coding.c
--- a/src/file-coding.c Sun Dec 22 10:36:33 2013 +0000
+++ b/src/file-coding.c Thu Jan 16 16:27:52 2014 +0000
@@ -1990,6 +1990,14 @@
return Lstream_seekable_p (str->other_end);
}
+static Charcount
+coding_character_tell (Lstream *stream)
+{
+ struct coding_stream *str = CODING_STREAM_DATA (stream);
+
+ return XCODESYSMETH_OR_GIVEN (str->codesys, character_tell, (str), -1);
+}
+
static int
coding_flusher (Lstream *stream)
{
@@ -2823,7 +2831,32 @@
#### Shouldn't we _call_ it that, then? And while we're at it,
separate it into "to_internal" and "to_external"? */
-DEFINE_CODING_SYSTEM_TYPE (no_conversion);
+
+
+struct no_conversion_coding_system
+{
+};
+
+struct no_conversion_coding_stream
+{
+ /* Number of characters seen when decoding. */
+ Charcount characters_seen;
+};
+
+static const struct memory_description
no_conversion_coding_system_description[] = {
+ { XD_END }
+};
+
+static const struct memory_description
no_conversion_coding_stream_description_1 [] = {
+ { XD_INT, offsetof (struct no_conversion_coding_stream, characters_seen) },
+ { XD_END }
+};
+
+const struct sized_memory_description no_conversion_coding_stream_description
= {
+ sizeof (struct no_conversion_coding_stream),
no_conversion_coding_stream_description_1
+};
+
+DEFINE_CODING_SYSTEM_TYPE_WITH_DATA (no_conversion);
/* This is used when reading in "binary" files -- i.e. files that may
contain all 256 possible byte values and that are not to be
@@ -2846,6 +2879,9 @@
DECODE_ADD_BINARY_CHAR (c, dst);
}
+ CODING_STREAM_TYPE_DATA (str, no_conversion)->characters_seen
+ += orign;
+
if (str->eof)
DECODE_OUTPUT_PARTIAL_CHAR (ch, dst);
}
@@ -2904,6 +2940,12 @@
return orign;
}
+static Charcount
+no_conversion_character_tell (struct coding_stream *str)
+{
+ return CODING_STREAM_TYPE_DATA (str, no_conversion)->characters_seen;
+}
+
DEFINE_DETECTOR (no_conversion);
DEFINE_DETECTOR_CATEGORY (no_conversion, no_conversion);
@@ -4656,6 +4698,7 @@
LSTREAM_HAS_METHOD (coding, writer);
LSTREAM_HAS_METHOD (coding, rewinder);
LSTREAM_HAS_METHOD (coding, seekable_p);
+ LSTREAM_HAS_METHOD (coding, character_tell);
LSTREAM_HAS_METHOD (coding, marker);
LSTREAM_HAS_METHOD (coding, flusher);
LSTREAM_HAS_METHOD (coding, closer);
@@ -4697,9 +4740,10 @@
dump_add_opaque_int (&coding_detector_count);
dump_add_opaque_int (&coding_detector_category_count);
- INITIALIZE_CODING_SYSTEM_TYPE (no_conversion,
- "no-conversion-coding-system-p");
+ INITIALIZE_CODING_SYSTEM_TYPE_WITH_DATA (no_conversion,
+ "no-conversion-coding-system-p");
CODING_SYSTEM_HAS_METHOD (no_conversion, convert);
+ CODING_SYSTEM_HAS_METHOD (no_conversion, character_tell);
INITIALIZE_DETECTOR (no_conversion);
DETECTOR_HAS_METHOD (no_conversion, detect);
diff -r 4004c3266c09 -r 65d65b52d608 src/file-coding.h
--- a/src/file-coding.h Sun Dec 22 10:36:33 2013 +0000
+++ b/src/file-coding.h Thu Jan 16 16:27:52 2014 +0000
@@ -353,6 +353,9 @@
a result of the stream being rewound. Optional. */
void (*rewind_coding_stream_method) (struct coding_stream *str);
+ /* Return the number of characters *decoded*. Optional. */
+ Charcount (*character_tell_method) (struct coding_stream *str);
+
/* Finalize coding stream method: Clean up the type-specific data
attached to the coding stream (i.e. in struct TYPE_coding_stream).
Happens when the Lstream is deleted using Lstream_delete() or is
diff -r 4004c3266c09 -r 65d65b52d608 src/fileio.c
--- a/src/fileio.c Sun Dec 22 10:36:33 2013 +0000
+++ b/src/fileio.c Thu Jan 16 16:27:52 2014 +0000
@@ -3180,6 +3180,7 @@
struct gcpro ngcpro1;
Lisp_Object stream = make_filedesc_input_stream (fd, 0, total,
LSTR_ALLOW_QUIT);
+ Charcount last_tell = -1;
NGCPRO1 (stream);
Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
@@ -3187,6 +3188,7 @@
(XLSTREAM (stream), get_coding_system_for_text_file (codesys, 1),
CODING_DECODE, 0);
Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+ last_tell = Lstream_character_tell (XLSTREAM (stream));
record_unwind_protect (delete_stream_unwind, stream);
@@ -3196,7 +3198,7 @@
while (1)
{
Bytecount this_len;
- Charcount cc_inserted;
+ Charcount cc_inserted, this_tell = last_tell;
QUIT;
this_len = Lstream_read (XLSTREAM (stream), read_buf,
@@ -3209,12 +3211,17 @@
break;
}
- cc_inserted = buffer_insert_raw_string_1 (buf, cur_point, read_buf,
- this_len,
- !NILP (visit)
- ? INSDEL_NO_LOCKING : 0);
+ cc_inserted
+ = buffer_insert_string_1 (buf, cur_point, read_buf, Qnil,
+ 0, this_len, last_tell >= 0
+ ? (this_tell
+ = Lstream_character_tell (XLSTREAM
+ (stream)))
+ - last_tell : -1,
+ !NILP (visit) ? INSDEL_NO_LOCKING : 0);
inserted += cc_inserted;
cur_point += cc_inserted;
+ last_tell = this_tell;
}
if (!NILP (used_codesys))
{
diff -r 4004c3266c09 -r 65d65b52d608 src/insdel.c
--- a/src/insdel.c Sun Dec 22 10:36:33 2013 +0000
+++ b/src/insdel.c Thu Jan 16 16:27:52 2014 +0000
@@ -1039,14 +1039,15 @@
#endif
}
-/* Insert a string into BUF at Charbpos POS. The string data comes
- from one of two sources: constant, non-relocatable data (specified
- in NONRELOC), or a Lisp string object (specified in RELOC), which
- is relocatable and may have extent data that needs to be copied
- into the buffer. OFFSET and LENGTH specify the substring of the
- data that is actually to be inserted. As a special case, if POS
- is -1, insert the string at point and move point to the end of the
- string.
+/* Insert a string into BUF at Charbpos POS. The string data comes from one
+ of two sources: constant, non-relocatable data (specified in NONRELOC),
+ or a Lisp string object (specified in RELOC), which is relocatable and
+ may have extent data that needs to be copied into the buffer. OFFSET and
+ LENGTH specify the substring of the data that is actually to be inserted.
+ As a special case, if POS is -1, insert the string at point and move
+ point to the end of the string. CCLEN is the character count of the data
+ to be inserted, and can be -1 to indicate that buffer_insert_string_1 ()
+ should work this out itself with bytecount_to_charcount().
Normally, markers at the insertion point end up before the
inserted string. If INSDEL_BEFORE_MARKERS is set in flags, however,
@@ -1061,13 +1062,12 @@
buffer_insert_string_1 (struct buffer *buf, Charbpos pos,
const Ibyte *nonreloc, Lisp_Object reloc,
Bytecount offset, Bytecount length,
- int flags)
+ Charcount cclen, int flags)
{
/* This function can GC */
struct gcpro gcpro1;
Bytebpos bytepos;
Bytecount length_in_buffer;
- Charcount cclen;
int move_point = 0;
struct buffer *mbuf;
Lisp_Object bufcons;
@@ -1118,14 +1118,27 @@
bytepos = charbpos_to_bytebpos (buf, pos);
- /* string may have been relocated up to this point */
- if (STRINGP (reloc))
+ if (cclen < 0)
{
- cclen = string_offset_byte_to_char_len (reloc, offset, length);
- nonreloc = XSTRING_DATA (reloc);
+ /* string may have been relocated up to this point */
+ if (STRINGP (reloc))
+ {
+ cclen = string_offset_byte_to_char_len (reloc, offset, length);
+ nonreloc = XSTRING_DATA (reloc);
+ }
+ else
+ cclen = bytecount_to_charcount (nonreloc + offset, length);
}
else
- cclen = bytecount_to_charcount (nonreloc + offset, length);
+ {
+ text_checking_assert (cclen > 0 && cclen
+ == (STRINGP (reloc) ?
+ string_offset_byte_to_char_len (reloc, offset,
+ length)
+ : bytecount_to_charcount (nonreloc + offset,
+ length)));
+ }
+
/* &&#### Here we check if the text can't fit into the format of the buffer,
and if so convert it to another format (either default or 32-bit-fixed,
according to some flag; if no flag, use default). */
@@ -1286,7 +1299,7 @@
{
/* This function can GC */
return buffer_insert_string_1 (buf, pos, nonreloc, Qnil, 0, length,
- flags);
+ -1, flags);
}
Charcount
@@ -1295,8 +1308,7 @@
{
/* This function can GC */
return buffer_insert_string_1 (buf, pos, 0, str, 0,
- XSTRING_LENGTH (str),
- flags);
+ XSTRING_LENGTH (str), -1, flags);
}
/* Insert the null-terminated string S (in external format). */
@@ -1309,7 +1321,7 @@
const CIbyte *translated = GETTEXT (s);
ASSERT_ASCTEXT_ASCII (s);
return buffer_insert_string_1 (buf, pos, (const Ibyte *) translated, Qnil,
- 0, strlen (translated), flags);
+ 0, strlen (translated), -1, flags);
}
Charcount
@@ -1319,7 +1331,7 @@
/* This function can GC */
Ibyte str[MAX_ICHAR_LEN];
Bytecount len = set_itext_ichar (str, ch);
- return buffer_insert_string_1 (buf, pos, str, Qnil, 0, len, flags);
+ return buffer_insert_string_1 (buf, pos, str, Qnil, 0, len, -1, flags);
}
Charcount
@@ -1339,7 +1351,7 @@
/* This function can GC */
Lisp_Object str = make_string_from_buffer (buf2, pos2, length);
return buffer_insert_string_1 (buf, pos, 0, str, 0,
- XSTRING_LENGTH (str), flags);
+ XSTRING_LENGTH (str), -1, flags);
}
@@ -1674,7 +1686,7 @@
* backward so that it now equals the insertion point.
*/
buffer_insert_string_1 (buf, (movepoint ? -1 : pos),
- newstr, Qnil, 0, newlen, 0);
+ newstr, Qnil, 0, newlen, -1, 0);
}
}
diff -r 4004c3266c09 -r 65d65b52d608 src/insdel.h
--- a/src/insdel.h Sun Dec 22 10:36:33 2013 +0000
+++ b/src/insdel.h Thu Jan 16 16:27:52 2014 +0000
@@ -38,7 +38,7 @@
Charcount buffer_insert_string_1 (struct buffer *buf, Charbpos pos,
const Ibyte *nonreloc, Lisp_Object reloc,
Bytecount offset, Bytecount length,
- int flags);
+ Charcount clen, int flags);
Charcount buffer_insert_raw_string_1 (struct buffer *buf, Charbpos pos,
const Ibyte *nonreloc,
Bytecount length, int flags);
@@ -58,7 +58,7 @@
All of these can GC. */
#define buffer_insert_string(buf, nonreloc, reloc, offset, length) \
- buffer_insert_string_1 (buf, -1, nonreloc, reloc, offset, length, 0)
+ buffer_insert_string_1 (buf, -1, nonreloc, reloc, offset, length, -1, 0)
#define buffer_insert_raw_string(buf, string, length) \
buffer_insert_raw_string_1 (buf, -1, string, length, 0)
#define buffer_insert_ascstring(buf, s) \
diff -r 4004c3266c09 -r 65d65b52d608 src/lstream.c
--- a/src/lstream.c Sun Dec 22 10:36:33 2013 +0000
+++ b/src/lstream.c Thu Jan 16 16:27:52 2014 +0000
@@ -735,6 +735,134 @@
return Lstream_read_1 (lstr, data, size, 0);
}
+Charcount
+Lstream_character_tell (Lstream *lstr)
+{
+ Charcount ctell = lstr->imp->character_tell ?
+ lstr->imp->character_tell (lstr) : -1;
+
+ if (ctell >= 0)
+ {
+ /* Our implementation's character tell code doesn't know about the
+ unget buffer, update its figure to reflect it. */
+ ctell += lstr->unget_character_count;
+
+ if (lstr->unget_buffer_ind > 0)
+ {
+ /* The character count should not include those characters
+ currently *in* the unget buffer, subtract that count. */
+ Ibyte *ungot, *ungot_ptr;
+ Bytecount ii = lstr->unget_buffer_ind, impartial, sevenflen;
+
+ ungot_ptr = ungot
+ = alloca_ibytes (lstr->unget_buffer_ind) + MAX_ICHAR_LEN;
+
+ /* Make sure the string starts with a valid ibyteptr, otherwise
+ validate_ibyte_string_backward could run off the beginning. */
+ sevenflen = set_itext_ichar (ungot, (Ichar) 0x7f);
+ ungot_ptr += sevenflen;
+
+ /* Internal format data, but in reverse order. There's not
+ actually a need to alloca here, we could work out the character
+ count directly from the reversed bytes, but the alloca approach
+ is more robust to changes in our internal format, and the unget
+ buffer is not going to blow the stack. */
+ while (ii > 0)
+ {
+ *ungot_ptr++ = lstr->unget_buffer[--ii];
+ }
+
+ impartial
+ = validate_ibyte_string_backward (ungot, ungot_ptr - ungot);
+
+ /* Move past the character we added. */
+ impartial -= sevenflen;
+ INC_IBYTEPTR (ungot);
+
+ if (impartial > 0 && !valid_ibyteptr_p (ungot))
+ {
+ Ibyte *newstart = ungot, *limit = ungot + impartial;
+ /* Our consumer has the start of a partial character, we
+ have the rest. */
+
+ while (!valid_ibyteptr_p (newstart) && newstart < limit)
+ {
+ newstart++, impartial--;
+ }
+
+ /* Remove this character from the count, since the
+ end-consumer hasn't seen the full character. */
+ ctell--;
+ ungot = newstart;
+ }
+ else if (valid_ibyteptr_p (ungot)
+ && rep_bytes_by_first_byte (*ungot) > impartial)
+ {
+ /* Rest of a partial character has yet to be read, its first
+ octet has probably been unread by Lstream_read_1(). We
+ included it in the accounting in Lstream_unread(), adjust
+ the figure here appropriately. */
+ ctell--;
+ }
+
+ /* bytecount_to_charcount will throw an assertion failure if we're
+ not at the start of a character. */
+ text_checking_assert (impartial == 0 || valid_ibyteptr_p (ungot));
+
+ /* The character length of this text is included in
+ unget_character_count; if the bytes are still in the unget
+ buffer, then our consumers haven't seen them, and so the
+ character tell figure shouldn't reflect them. Subtract it from
+ the total. */
+ ctell -= bytecount_to_charcount (ungot, impartial);
+ }
+
+ if (lstr->in_buffer_ind < lstr->in_buffer_current)
+ {
+ Ibyte *inbuf = lstr->in_buffer + lstr->in_buffer_ind;
+ Bytecount partial = lstr->in_buffer_current - lstr->in_buffer_ind,
+ impartial;
+
+ if (!valid_ibyteptr_p (inbuf))
+ {
+ Ibyte *newstart = inbuf;
+ Ibyte *limit = lstr->in_buffer + lstr->in_buffer_current;
+ /* Our consumer has the start of a partial character, we
+ have the rest. */
+
+ while (newstart < limit && !valid_ibyteptr_p (newstart))
+ {
+ newstart++;
+ }
+
+ /* Remove this character from the count, since the
+ end-consumer hasn't seen the full character. */
+ ctell--;
+ inbuf = newstart;
+ partial = limit - newstart;
+ }
+
+ if (valid_ibyteptr_p (inbuf))
+ {
+ /* There's at least one valid starting char in the string,
+ validate_ibyte_string_backward won't run off the
+ begining. */
+ impartial =
+ validate_ibyte_string_backward (inbuf, partial);
+ }
+ else
+ {
+ impartial = 0;
+ }
+
+ ctell -= bytecount_to_charcount (inbuf, impartial);
+ }
+
+ text_checking_assert (ctell >= 0);
+ }
+
+ return ctell;
+}
/* Push back SIZE bytes of DATA onto the input queue. The next call
to Lstream_read() with the same size will read the same bytes back.
@@ -755,7 +883,12 @@
/* Bytes have to go on in reverse order -- they are reversed
again when read back. */
while (size--)
- lstr->unget_buffer[lstr->unget_buffer_ind++] = p[size];
+ {
+ lstr->unget_buffer[lstr->unget_buffer_ind++] = p[size];
+ /* If we see a valid first byte, that is the last octet in a
+ character, so increase the count of ungot characters. */
+ lstr->unget_character_count += valid_ibyteptr_p (p + size);
+ }
}
/* Rewind the stream to the beginning. */
@@ -768,6 +901,7 @@
if (Lstream_flush (lstr) < 0)
return -1;
lstr->byte_count = 0;
+ lstr->unget_character_count = 0;
return (lstr->imp->rewinder) (lstr);
}
diff -r 4004c3266c09 -r 65d65b52d608 src/lstream.h
--- a/src/lstream.h Sun Dec 22 10:36:33 2013 +0000
+++ b/src/lstream.h Thu Jan 16 16:27:52 2014 +0000
@@ -181,6 +181,10 @@
method. If this method is not present, the result is determined
by whether a rewind method is present. */
int (*seekable_p) (Lstream *stream);
+
+ /* Return the number of complete characters read so far. Respects
+ buffering and unget. Returns -1 if unknown or not implemented. */
+ Charcount (*character_tell) (Lstream *stream);
/* Perform any additional operations necessary to flush the
data in this stream. */
int (*flusher) (Lstream *stream);
@@ -250,8 +254,9 @@
similarly has to push the data on backwards. */
unsigned char *unget_buffer; /* holds characters pushed back onto input */
Bytecount unget_buffer_size; /* allocated size of buffer */
- Bytecount unget_buffer_ind; /* pointer to next buffer spot
- to write a character */
+ Bytecount unget_buffer_ind; /* Next buffer spot to write a character */
+
+ Charcount unget_character_count; /* Count of complete characters ever ungot.
*/
Bytecount byte_count;
int flags;
@@ -297,8 +302,8 @@
int Lstream_fputc (Lstream *lstr, int c);
int Lstream_fgetc (Lstream *lstr);
void Lstream_fungetc (Lstream *lstr, int c);
-Bytecount Lstream_read (Lstream *lstr, void *data,
- Bytecount size);
+Bytecount Lstream_read (Lstream *lstr, void *data, Bytecount size);
+Charcount Lstream_character_tell (Lstream *);
int Lstream_write (Lstream *lstr, const void *data,
Bytecount size);
int Lstream_was_blocked_p (Lstream *lstr);
@@ -353,19 +358,28 @@
reverse order they were pushed back -- most recent first. (This is
necessary for consistency -- if there are a number of bytes that
have been unread and I read and unread a byte, it needs to be the
- first to be read again.) This is a macro and so it is very
- efficient. The C argument is only evaluated once but the STREAM
- argument is evaluated more than once.
- */
+ first to be read again.) */
-#define Lstream_ungetc(stream, c) \
-/* Add to the end if it won't overflow buffer; otherwise call the \
- function equivalent */ \
- ((stream)->unget_buffer_ind >= (stream)->unget_buffer_size ? \
- Lstream_fungetc (stream, c) : \
- (void) ((stream)->byte_count--, \
- ((stream)->unget_buffer[(stream)->unget_buffer_ind++] = \
- (unsigned char) (c))))
+DECLARE_INLINE_HEADER (
+void
+Lstream_ungetc (Lstream *lstr, int c)
+)
+{
+ /* Add to the end if it won't overflow buffer; otherwise call the
+ function equivalent */
+ if (lstr->unget_buffer_ind >= lstr->unget_buffer_size)
+ {
+ Lstream_fungetc (lstr, c);
+ }
+ else
+ {
+ lstr->byte_count--;
+ lstr->unget_buffer[lstr->unget_buffer_ind] = (unsigned char) (c);
+ lstr->unget_character_count
+ += valid_ibyteptr_p (lstr->unget_buffer + lstr->unget_buffer_ind);
+ lstr->unget_buffer_ind++;
+ }
+}
#define Lstream_data(stream) ((void *) ((stream)->data))
#define Lstream_byte_count(stream) ((stream)->byte_count)
diff -r 4004c3266c09 -r 65d65b52d608 src/print.c
--- a/src/print.c Sun Dec 22 10:36:33 2013 +0000
+++ b/src/print.c Thu Jan 16 16:27:52 2014 +0000
@@ -514,7 +514,7 @@
buffer_insert_string_1 (XMARKER (function)->buffer,
spoint, nonreloc, reloc, offset, len,
- 0);
+ -1, 0);
Fset_marker (function, make_fixnum (spoint + cclen),
Fmarker_buffer (function));
}
diff -r 4004c3266c09 -r 65d65b52d608 src/tests.c
--- a/src/tests.c Sun Dec 22 10:36:33 2013 +0000
+++ b/src/tests.c Thu Jan 16 16:27:52 2014 +0000
@@ -558,6 +558,186 @@
return conversion_result;
}
+DEFUN ("test-character-tell", Ftest_character_tell, 0, 0, "", /*
+Return list of results of tests of the stream character offset code.
+For use by the automated test suite. See tests/automated/c-tests.
+
+Each element is a list (DESCRIPTION, STATUS, REASON).
+DESCRIPTION is a string describing the test.
+STATUS is a symbol, either t (pass) or nil (fail).
+REASON is nil or a string describing the failure (not required).
+*/
+ ())
+{
+ Extbyte ext_unix[]= "\n\nfoo\nbar\n\nf\372b\343\340\nfoo\nbar\n";
+ /* Previous string in UTF-8. */
+ Extbyte ext_utf_8_unix[]
+ = "\n\nfoo\nbar\n\nf\303\272b\303\243\303\240\nfoo\nbar\n";
+ Charcount ext_utf_8_unix_char_len = 25;
+ Ibyte shortbuf[13], longbuf[512];
+ Lisp_Object stream =
+ make_fixed_buffer_input_stream (ext_unix, sizeof (ext_unix) - 1);
+ Lisp_Object result = Qnil, string = Qnil;
+ Charcount count;
+ Bytecount bytecount;
+ struct gcpro gcpro1, gcpro2, gcpro3;
+
+#define CHARACTER_TELL_ASSERT(assertion, description, failing_case) \
+ do \
+ { \
+ if (assertion) \
+ result = Fcons (list3 (build_cistring (description), \
+ Qt, Qnil), result); \
+ else \
+ result = Fcons (list3 (build_cistring (description), \
+ Qnil, build_ascstring (failing_case)), \
+ result); \
+ } \
+ while (0)
+
+ GCPRO3 (stream, result, string);
+
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+ stream = make_coding_input_stream
+ (XLSTREAM (stream), Ffind_coding_system (intern ("no-conversion-unix")),
+ CODING_DECODE, 0);
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+
+ bytecount = Lstream_read (XLSTREAM (stream), longbuf, sizeof (longbuf));
+
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == sizeof (ext_unix) -1,
+ "basic character tell, no-conversion-unix",
+ "basic character tell failed");
+
+ string = build_extstring (ext_unix,
+ Ffind_coding_system (intern
+ ("no-conversion-unix")));
+
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == string_char_length (string),
+ "repeat basic character tell, no-conversion-unix",
+ "repeat basic character tell failed with string");
+
+ count = Lstream_character_tell (XLSTREAM (stream));
+
+ Lstream_unread (XLSTREAM (stream), "r\n", 2);
+
+ /* This should give the same result as before the unread. */
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == count, "checking post-unread character tell",
+ "post-unread character tell failed");
+ bytecount += Lstream_read (XLSTREAM (stream), longbuf + bytecount,
+ sizeof (longbuf) - bytecount);
+
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == count + 2,
+ "checking post-unread+read character tell",
+ "post-unread+read character tell failed");
+
+ /* This seems to be buggy for my purposes. */
+ /* Lstream_rewind (XLSTREAM (stream)); */
+ Lstream_close (XLSTREAM (stream));
+ Lstream_delete (XLSTREAM (stream));
+
+ stream = make_fixed_buffer_input_stream (ext_unix, sizeof (ext_unix) - 1);
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+ Lstream_unset_character_mode (XLSTREAM (stream));
+ stream = make_coding_input_stream
+ (XLSTREAM (stream), Ffind_coding_system (intern ("no-conversion-unix")),
+ CODING_DECODE, 0);
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+ Lstream_unset_character_mode (XLSTREAM (stream));
+
+ bytecount = Lstream_read (XLSTREAM (stream), shortbuf, sizeof (shortbuf));
+
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ /* This should be equal to sizeof (shortbuf) on
+ non-mule. */
+ == sizeof (shortbuf) - !(byte_ascii_p (0xff)),
+ "character tell with short read, no-conversion-unix",
+ "short read character tell failed");
+
+ Lstream_close (XLSTREAM (stream));
+ Lstream_delete (XLSTREAM (stream));
+
+ stream
+ = make_fixed_buffer_input_stream (ext_utf_8_unix,
+ sizeof (ext_utf_8_unix) - 1);
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+ stream = make_coding_input_stream
+ (XLSTREAM (stream), Ffind_coding_system (intern ("utf-8-unix")),
+ CODING_DECODE, 0);
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+
+ bytecount = Lstream_read (XLSTREAM (stream), longbuf, sizeof (longbuf));
+
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == ext_utf_8_unix_char_len,
+ "utf-8 character tell, utf-8-unix",
+ "utf-8 character tell failed");
+
+ string = build_extstring (ext_utf_8_unix,
+ Ffind_coding_system (intern
+ ("utf-8-unix")));
+
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == string_char_length (string),
+ "repeat utf-8 character tell, utf-8-unix",
+ "repeat utf-8 character tell failed with string");
+
+ count = Lstream_character_tell (XLSTREAM (stream));
+
+ Lstream_unread (XLSTREAM (stream), "r\n", 2);
+
+ /* This should give the same result as before the unread. */
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == count, "checking post-unread utf-8 tell",
+ "post-unread utf-8 tell failed");
+ bytecount += Lstream_read (XLSTREAM (stream), longbuf + bytecount,
+ sizeof (longbuf) - bytecount);
+
+ CHARACTER_TELL_ASSERT (Lstream_character_tell (XLSTREAM (stream))
+ == count + 2,
+ "checking post-unread+read utf-8 tell",
+ "post-unread+read utf-8 tell failed");
+
+ /* This seems to be buggy for my purposes. */
+ /* Lstream_rewind (XLSTREAM (stream)); */
+ Lstream_close (XLSTREAM (stream));
+ Lstream_delete (XLSTREAM (stream));
+
+ stream = make_fixed_buffer_input_stream (ext_utf_8_unix, sizeof
(ext_utf_8_unix) - 1);
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+ Lstream_set_character_mode (XLSTREAM (stream));
+
+ stream = make_coding_input_stream
+ (XLSTREAM (stream), Ffind_coding_system (intern ("utf-8-unix")),
+ CODING_DECODE, 0);
+ Lstream_set_buffering (XLSTREAM (stream), LSTREAM_BLOCKN_BUFFERED, 65536);
+ Lstream_set_character_mode (XLSTREAM (stream));
+
+ bytecount = Lstream_read (XLSTREAM (stream), shortbuf, sizeof (shortbuf));
+
+ CHARACTER_TELL_ASSERT
+ (bytecount == (sizeof (shortbuf) - 1),
+ "utf-8 Lstream_read, character mode, checking partial char not read",
+ "partial char appars to have been read when it shouldn't");
+
+ CHARACTER_TELL_ASSERT
+ (Lstream_character_tell (XLSTREAM (stream))
+ /* This is shorter, because it's in the middle of a character. */
+ == sizeof (shortbuf) - 1,
+ "utf-8 tell with short read, character mode, utf-8-unix",
+ "utf-8 read character tell, character mode failed");
+
+ Lstream_close (XLSTREAM (stream));
+ Lstream_delete (XLSTREAM (stream));
+
+ UNGCPRO;
+ return result;
+}
+
/* Hash Table testing */
@@ -724,6 +904,7 @@
Vtest_function_list = Qnil;
TESTS_DEFSUBR (Ftest_data_format_conversion);
+ TESTS_DEFSUBR (Ftest_character_tell);
TESTS_DEFSUBR (Ftest_hash_tables);
TESTS_DEFSUBR (Ftest_store_void_in_lisp);
/* Add other test functions here with TESTS_DEFSUBR */
diff -r 4004c3266c09 -r 65d65b52d608 src/unicode.c
--- a/src/unicode.c Sun Dec 22 10:36:33 2013 +0000
+++ b/src/unicode.c Thu Jan 16 16:27:52 2014 +0000
@@ -1707,6 +1707,7 @@
unsigned char counter;
unsigned char indicated_length;
int seen_char;
+ Charcount characters_seen;
/* encode */
Lisp_Object current_charset;
int current_char_boundary;
@@ -1988,6 +1989,17 @@
write_error_characters_as_such);
}
+static Charcount
+unicode_character_tell (struct coding_stream *str)
+{
+ if (CODING_STREAM_TYPE_DATA (str, unicode)->counter == 0)
+ {
+ return CODING_STREAM_TYPE_DATA (str, unicode)->characters_seen;
+ }
+
+ return -1;
+}
+
static Bytecount
unicode_convert (struct coding_stream *str, const UExtbyte *src,
unsigned_char_dynarr *dst, Bytecount n)
@@ -2006,6 +2018,7 @@
unsigned char counter = data->counter;
unsigned char indicated_length
= data->indicated_length;
+ Charcount characters_seen = data->characters_seen;
while (n--)
{
@@ -2020,12 +2033,15 @@
{
/* ASCII. */
decode_unicode_char (c, dst, data, ignore_bom);
+ characters_seen++;
}
else if (0 == (c & 0x40))
{
/* Highest bit set, second highest not--there's
something wrong. */
DECODE_ERROR_OCTET (c, dst, data, ignore_bom);
+ /* This is a character in the buffer. */
+ characters_seen++;
}
else if (0 == (c & 0x20))
{
@@ -2050,7 +2066,7 @@
/* We don't supports lengths longer than 4 in
external-format data. */
DECODE_ERROR_OCTET (c, dst, data, ignore_bom);
-
+ characters_seen++;
}
}
else
@@ -2061,15 +2077,20 @@
indicate_invalid_utf_8(indicated_length,
counter,
ch, dst, data, ignore_bom);
+ /* These are characters our receiver will see, not
+ actual characters we've seen in the input. */
+ characters_seen += (indicated_length - counter);
if (c & 0x80)
{
DECODE_ERROR_OCTET (c, dst, data, ignore_bom);
+ characters_seen++;
}
else
{
/* The character just read is ASCII. Treat it as
such. */
decode_unicode_char (c, dst, data, ignore_bom);
+ characters_seen++;
}
ch = 0;
counter = 0;
@@ -2092,10 +2113,12 @@
counter,
ch, dst, data,
ignore_bom);
+ characters_seen += (indicated_length - counter);
}
else
{
decode_unicode_char (ch, dst, data, ignore_bom);
+ characters_seen++;
}
ch = 0;
}
@@ -2242,6 +2265,7 @@
indicate_invalid_utf_8(indicated_length,
counter, ch, dst, data,
ignore_bom);
+ characters_seen += (indicated_length - counter);
break;
case UNICODE_UTF_16:
@@ -2295,6 +2319,7 @@
data->counter = counter;
data->indicated_length = indicated_length;
+ data->characters_seen = characters_seen;
}
else
{
@@ -3177,6 +3202,8 @@
CODING_SYSTEM_HAS_METHOD (unicode, putprop);
CODING_SYSTEM_HAS_METHOD (unicode, getprop);
+ CODING_SYSTEM_HAS_METHOD (unicode, character_tell);
+
INITIALIZE_DETECTOR (utf_8);
DETECTOR_HAS_METHOD (utf_8, detect);
INITIALIZE_DETECTOR_CATEGORY (utf_8, utf_8);
--
‘Liston operated so fast that he once accidentally amputated an assistant’s
fingers along with a patient’s leg, […] The patient and the assistant both
died of sepsis, and a spectator reportedly died of shock, resulting in the
only known procedure with a 300% mortality.’ (Atul Gawande, NEJM, 2012)
| [Prev in Thread] |
Current Thread |
[Next in Thread] |
- [VM] [COMMIT XEMACS] Pass character count from coding systems to buffer insertion code.,
Aidan Kehoe <=