Changes Log for html2text

Changes in Version 1.3.2(a) (released 2004-01-15)

Bugfixes and new features

Version 1.3.2a only: ported to gcc 3.3+

Ported to g++ v. 3.3. Since this change is not compatible to g++ versions prior to 3.0, it is included in version 1.3.2A only. THIS IS THE ONLY DIFFERENCE BETWEEN VERSIONS 1.2.3 AND 1.2.3A. (Area.C, Area.h, HTMLControl.C, HTMLControl.h, Properties.C, Properties.h, format.C, format.h, html.C, html.h, html2text.C, urlistream.h)

Added command line option '-ascii' for straight ascii output (instead of ISO-8859-1, which is the default). (html2text.C, sgml.C)

Implemented rendering of most SGML entities introduced in HTML-4. As a limitation, those entities not present in HTML-3.2/ISO-8859-1 will be recognized only if represented as "named entities" and not with thier numeric values, e.g. "™" will be rendered, "™" won't. (html2text.C, sgml.C)

Element closing as done in XHTML (e.g. "<br />") is now tolerated. (HTMLControl.C)

The program now ignores the content of <SCRIPT> or <STYLE> elemts within TABLEs, even if it is not commented out. (HTMLParser.y)

Fixed urlistream.h (fd_ might be uninitialized). (urlistream.h)

Changes in Version 1.3.1 (released 2002-09-02)

Bugfix release

Multiple-line DOCTYPE declarations are now accepted (HTMLControl.C:352).

Bad initialisation of "fd_" in "urlistream" fixed (urlistream.h), so that an error message is printed if a remote document could not be retrieved.

A missing node now is interpreted as node '/' (urlistream.C).

Closing DIV, FORM and BLOCKQUOTE tags as well as the closing UL, OL and PRE tags are now optional (HTMLParser.y), so that the program won't die any longer if one of them is omited in the document.

Some dificulties with non-ASCII chars fixed (HTMLControl.C).

Colons in elements and attributes now are tolerated and elements are not any longer implecitly closed at a newline (HTMLControl.C), in order to handly MS-Word's HTML better.

We do not use '/dev/stdin' as file descriptor any longer (html2text.C and urlistream.C), in order to make reading from STDIN finaly work.

Block elements are not enclosed in -implecit- Paragraphs any longer (HTMLParser.y), in order to avoid superfluous newlines in the output.

Fixed segmentation fault on tables with "border" attribute: we now assume that any TABLE has at least one row and one column (table.C).

Fixed format.C to avoid excessive runtime increment on parsing much nested block elements.

Some changes in configure, and the documentation.

Changes in Version 1.3.0 (released 2001-10-08)

Ported to g++-3.0 and GPL

Ported to g++ version 3.0. This uses the 'istream.h' header file from the g++3's 'backward' directory.

Bugfix: '-' did not work as synonym for STDIN.

Added support for the EURO-sign (well, almost).

Finaly the GNU GPL as new copyright terms for all parts of the program, after GMRS agreed to change the program's license terms to it.

Changes in Version 1.2.4 (released 2001-06-11)

Bugfix and New Features

Fixed coredump when parsing empty <SCRIPT> or <STYLE> elements (HTMLControl.C).

New image handling: <IMG alt=""> does no longer return the value of the SRC attribute nor "[]". Added new ~/.html2textrc options: IMG.replace.{all noalt} and IMG.alt.{prefix suffix} with new defaults in pretty-style mode. Added method for checking whether an attribut was set, even to a zero-value.

(New) Copyright terms for all changes we made since version 1.2.2.

Some minor changes in configure and html2text.C.

Updated the documentation.

Changes in Version 1.2.3 (released 2001-03-19)

Bugfix Release

Fixed segmention fault when parsing <H5> tags (typo in format.C).

Changes in Version 1.2.2a (released 2000-09-15)

New Location

No changes were made. We call it now Version 1.2.2a only in order to make it perfectly clear to anybody that this is no longer the package deriving straight from GMRS, since they do not maintain this program any longer.

Changes in Version 1.2.2

let "make clean" remove BISON++-generated file only when BISON++ is installed.

Fixed a nasty one-line bug...

Changes in Version 1.2.1

Closing comment with triple dash did not work ("Ilya Kheifets" <ihf[at]agava[dot]com>).

Fixed up "depend" / "clean" / "clobber" rule.

Improved the "MAKEDEPEND_INCLUDES" and "SOCKET_LIBRARIES" algorithm (raichle[at]informatik[dot]uni-stuttgart[dot]de).

Fixed up "-help". Fixed "getenv("HOME")" core dump (bubeck[at]think-at-work[dot]de).

Solaris needs "#include <sys/types.h>" (raichle[at]informatik[dot]uni-stuttgart[dot]de).

Fixed up URLs with "?" (bubeck[at]think-at-work[dot]de).

Changes in Version 1.2

Added "$Id: ChangeLog,v 1.10 2000/02/28 18:42:47 arno Exp $" to files wher it was missing.

Added "Formatting::setProperty()".

Added "DL.vspace{before between after}".

Added "Formatting::setProperty()".

Added the "-style pretty" command line option.

Documented the "-style pretty" command line option.

Added missing implementation of "rb_tree:swap()".

Changes in Version 1.1.1

Removed the "STRING_ERASE" crap -- don't use G++ Standard C++ Library any more!


Added HAS_SYS_POLL, because Linux 2.0.35 does not provide <sys/poll.h>.

Removed the "tar-file" target and added a "make-tar-file" shell script.

CONFIGURE now uses $CXX instead of "cc" to check <sys/poll.h> and the socket libraries. The TMP dir is no longer "/tmp", but "./configure-tmp".

Replaced "echo -n" with "echo xxx\c". G++ 2.95.1 has problems with auto_ptrs and lists (see "configure").

Bug: Narrow table cells were not formatted properly.

Default format for <MENU> is now NO_BULLET / indent 2. Corrected indentation for <DT> / <DD>.

Added VSPACEs/INDENTs for <DT> and <DD>.

Fixed a formatting bug concerning <PRE> (Rolf Niepraschk, Nov 8 1999, 12:51).

Performance optimization: Do temporary table cell formatting always left aligned.

G++ fails to compile "multimap::equal_range()" for some unknown reason -- commented out.

Need a dummy "value_type" typedef for "vector" -- else G++ fails to compile.

Changes in Version 1.1

The formatting of HTML2TEXT is now largely customizable through the "$HOME/.html2textrc" file.

Inlined "Area::append()".

Optimized "Area::operator>>=()".

Formatting is now (partly) customizable, as proposed by Rolf Niepraschk. Added some missing copyright notices.

The default <UL> / <DIR> / <MENU> bullet style now depends on the list nesting l evel.

"Properties" now handles escape sequences.

Added "cmp_nocase()".

Added multiple-inclusion protection code. Added "cmp_nocase()".

Added new formatting properties.

Empty "HR.marker" now maps to a blank marker. Do not break lines at ",.;:". Break lines at "([{".

Added three custom bullet types.

niepraschk[at]ptb[dot]de: HTML2TEXT hangs when it encounters a "<TD COLSPAN=99>", because for HTML2TEXT table columns are separated by blanks, and 99 blanks are wider than 79 characters.

"urlistream" now strips "#anchor" suffixes from URLs.

niepraschk[at]ptb[dot]de: egcs-2.91.66 complains that "list::reverse_iterator" must not modify private "list::iterator::node".

Added man page for "html2textrc".

Changes in Version 1.04

Ported to EGCS 2.91.66 and SINIX CDS++ 2.00A00.

Improved generation of $MAKEDEPEND_INCLUDES: Now we use "$(CXX) -E" and look at the "#line" directives to find out where the C++ compiler found "#include <iostream.h>".

Removed a useless CONST qualifier in "libstd/include/string".

Cleaned up the make file.

"libstd/Makefile" is now only created when "./libstd" is actually needed.

Port to CDS++ 2.0: "string::operator[]()" is a bit picky... fixed.

Replaced "$(HAS_AUTO_PTR)" with "$(HAS_WORKING_AUTO_PTR)" -- "EGCS 2.91.66"'s implementation of "auto_ptr" is buggy!

Changes in Version 1.03

Version 1.03 provides fixes for two bugs reported by Volker Kuhlmann ("<MAP>" causes a CORE dump; "<A NAME=xxx>" causes the whole document being underlined), and also other fixes, as denoted below.

Also, attributes with textual values (like "ALIGN=RIGHT") were not processed case-insensitively.

Allow paragraph content in heading.

Removed "auto_aptr::operator const void *()" and "operator!()".

"Body::unparse()" and "Body::format()" were not virtual, so THs were unparsed and formatted as TDs.

Removed "auto_ptr<T>::operator=(T *)" for ANSI C++ compatibility. Non-inlined "HTMLParser::~HTMLParser()" to work around an EGCS bug.

Added "$EXPLICIT".

Removed "auto_ptr<T>::operator=(T *)" for ANSI C++ compatibility. Non-inlined "Element::~Element()" to work around an EGCS bug.

"Option.pcdata" is now an "auto_ptr" (memory leak).

Removed "auto_ptr<T>::operator=(T *)" for ANSI C++ compatibility.

Added "/usr/include/g++-2" to the MAKEDEPEND include path.

Added "cmp_nocase()". Swallow empty list items. Do not underline anchors that have no HREF.

The "get_attribute()" variants that compare attribute values against strings now compare case-insensitively. Added "cmp_nocase()".

Added multiple-inclusion protection code. Added "cmp_nocase()".

"Area::add_attribute()": Do not add attribute to left and right free areas.

Typo: "ALIGN" attribute of "<TR>" and "<TD>" was misspelled "HALIGN".

Changes in Version 1.02

Version 1.02 greatly improves the package's portability by adding a "configure" shell script.

date: Monday, June 21, 1999 @ 14:36

Modified Files: Makefile.Linux Makefile.SINIX html2text.1 html2text.C

Added Files: urlistream.C urlistream.h

Log Message: Source documents can now be retrieved over HTTP, not only from local files.

date: Thursday, June 24, 1999 @ 17:06

Added Files: configure

Removed Files: Makefile.Linux Makefile.SINIX

Log Message: Replaced "Makefile.SINIX" and "Makefile.Linux" with "" and "configure " -- improves portability!

Modified Files: format.C

Log Message: Changed the unchecked checkbox symbol from "O" to "º"

Modified Files: html2text.C

Log Message: Corrected the "-help" output (input URLs instead of input files).

Modified Files: sgml.C

Log Message: Eliminated the assumption that "string::operator[](x)" returns "eos" if "x==len"

Modified Files: urlistream.C urlistream.h

Log Message: Added the "timeout" parameter to "urlistream::open()". Added "urlistream::open_error()". Added a workaround for a SINIX OS bug.

date: Friday, July 2, 1999 @ 9:23

Modified Files: Area.h ChangeLog HTMLControl.C HTMLControl.h HTMLParser.y INSTALL configure format.C html.C sgml.C sgml.h table.C

Modified Files: auto_ptr.h list string

Removed Files: bool.h

Log Message: Added the "BOOL_DEFINITION" macro. Added the "STRING_ERASE" macro. Now allowing "-->" as the terminator of a multi-line comment. Removed "auto_ptr::operator void *()" and "auto_ptr::operator\!()" (portability) Updated the "INSTALL" file (usage of "sh configure" explained). Removed "libstd/include/bool.h" (was superseeded with "BOOL_DEFINITION"). Added some more MAKEDEPEND include directories. Added some temporary const reference variables to eliminate an incompatibility b etween "iterator" and "const_iterator" in some Standard C++ libraries. Cleaned up the SGML entity substitution code. Added "string::at()".

Changes in Version 1.01

Version 1.01 fixes a few bugs and improves the syntax error tolerance.

date: Thursday, June 17, 1999 @ 10:43

Modified Files: HTMLParser.y

Log Message: "</CENTER>" is now optional. Improved the syntax error tolerance (more "error" tokens).

Modified Files: Makefile.SINIX

Added Files: auto_aptr.h

Log Message: Added the "auto_aptr" template -- "auto_ptr"s for array pointers!

Modified Files: format.C

Log Message: "<INPUT>"'s "TYPE" field is now case-insensitive. Fixed a bug in the "<BR>" treatment of "make_up()".

Modified Files: table.C

Log Message: Bug: "column_widths" and "row_heights" were "static" -- problems with nested tables! (Now they are "auto_aptr"s.)

Modified Files: string Log Message: Added non-const "string::operator[]()".

date: Thursday, June 17, 1999 @ 14:46

Modified Files: HTMLControl.C

Log Message: Moved the SGML entity replacement from "HTMLControl.C" to "sgml.C". Accept empty tag attribute values.

Modified Files: Makefile.SINIX Makefile.Linux html2text.C Log Message: Version string is now compiled into "html2text -help" via "-DVERSION=...".

Modified Files: format.C

Log Message: The "ALT" tag attribute is now SGML-entity-expanded before it is displayed. "INPUT TYPE=TEXT" now displays the "NAME" tag attribute value if the "VALUE" tag attribute value is empty.

Modified Files: string.C

Log Message: Added one more "string::replace()" function. Added "string::traits_type" and its "eos()".

Modified Files: string

Log Message: Added one more "string::replace()" function. Added "string::traits_type" and its "eos()".

Added Files: sgml.C sgml.h

Log Message: Moved the SGML entity replacement from "HTMLControl.C" to "sgml.C".

date: Friday, June 18, 1999 @ 9:54

Modified Files: HTMLControl.C

Log Message: Allow underscores in tag names and tag attribute names.

Modified Files: HTMLParser.y

Log Message: Extension: Allow a preamble in a DL. DL more syntax error tolerant.

Modified Files: Makefile.SINIX

Log Message: Removed duplicate definition of "$(VERSION)". Target "tar-file" now checks if the version string in the man page is identical with $(VERSION).

Modified Files: format.C html.h

Log Message: Extension: Allow a preamble in a DL. DL term name is now indented by two characters.

Modified Files: Makefile.Linux Makefile.SINIX html2text.1 html2text.C

Log Message: Advanced from 1.0 to 1.01. Added "ChangeLog" to the distribution. Updated the manual page. Added the "-version" command line option.