gecko/netwerk/streamconv/converters/mozTXTToHTMLConv.h
Ehsan Akhgari 0fd9123eac Bug 579517 - Part 1: Automated conversion of NSPR numeric types to stdint types in Gecko; r=bsmedberg
This patch was generated by a script.  Here's the source of the script for
future reference:

function convert() {
echo "Converting $1 to $2..."
find . ! -wholename "*nsprpub*" \
       ! -wholename "*security/nss*" \
       ! -wholename "*/.hg*" \
       ! -wholename "obj-ff-dbg*" \
       ! -name nsXPCOMCID.h \
       ! -name prtypes.h \
         -type f \
      \( -iname "*.cpp" \
         -o -iname "*.h" \
         -o -iname "*.c" \
         -o -iname "*.cc" \
         -o -iname "*.idl" \
         -o -iname "*.ipdl" \
         -o -iname "*.ipdlh" \
         -o -iname "*.mm" \) | \
    xargs -n 1 sed -i -e "s/\b$1\b/$2/g"
}

convert PRInt8 int8_t
convert PRUint8 uint8_t
convert PRInt16 int16_t
convert PRUint16 uint16_t
convert PRInt32 int32_t
convert PRUint32 uint32_t
convert PRInt64 int64_t
convert PRUint64 uint64_t

convert PRIntn int
convert PRUintn unsigned

convert PRSize size_t

convert PROffset32 int32_t
convert PROffset64 int64_t

convert PRPtrdiff ptrdiff_t

convert PRFloat64 double
2012-08-22 11:56:38 -04:00

295 lines
12 KiB
C++

/* -*- Mode: C; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
/**
Description: Currently only functions to enhance plain text with HTML tags. See mozITXTToHTMLConv. Stream conversion is defunct.
*/
#ifndef _mozTXTToHTMLConv_h__
#define _mozTXTToHTMLConv_h__
#include "mozITXTToHTMLConv.h"
#include "nsIIOService.h"
#include "nsString.h"
#include "nsCOMPtr.h"
class mozTXTToHTMLConv : public mozITXTToHTMLConv
{
//////////////////////////////////////////////////////////
public:
//////////////////////////////////////////////////////////
mozTXTToHTMLConv();
virtual ~mozTXTToHTMLConv();
NS_DECL_ISUPPORTS
NS_DECL_MOZITXTTOHTMLCONV
NS_DECL_NSIREQUESTOBSERVER
NS_DECL_NSISTREAMLISTENER
NS_DECL_NSISTREAMCONVERTER
/**
see mozITXTToHTMLConv::ScanTXT
*/
void ScanTXT(const PRUnichar * aInString, int32_t aInStringLength, uint32_t whattodo, nsString& aOutString);
/**
see mozITXTToHTMLConv::ScanHTML. We will modify aInString potentially...
*/
void ScanHTML(nsString& aInString, uint32_t whattodo, nsString &aOutString);
/**
see mozITXTToHTMLConv::CiteLevelTXT
*/
int32_t CiteLevelTXT(const PRUnichar * line,uint32_t& logLineStart);
//////////////////////////////////////////////////////////
protected:
//////////////////////////////////////////////////////////
nsCOMPtr<nsIIOService> mIOService; // for performance reasons, cache the netwerk service...
/**
Completes<ul>
<li>Case 1: mailto: "mozilla@bucksch.org" -> "mailto:mozilla@bucksch.org"
<li>Case 2: http: "www.mozilla.org" -> "http://www.mozilla.org"
<li>Case 3: ftp: "ftp.mozilla.org" -> "ftp://www.mozilla.org"
</ul>
It does no check, if the resulting URL is valid.
@param text (in): abbreviated URL
@param pos (in): position of "@" (case 1) or first "." (case 2 and 3)
@return Completed URL at success and empty string at failure
*/
void CompleteAbbreviatedURL(const PRUnichar * aInString, int32_t aInLength,
const uint32_t pos, nsString& aOutString);
//////////////////////////////////////////////////////////
private:
//////////////////////////////////////////////////////////
enum LIMTYPE
{
LT_IGNORE, // limitation not checked
LT_DELIMITER, // not alphanumeric and not rep[0]. End of text is also ok.
LT_ALPHA, // alpha char
LT_DIGIT
};
/**
@param text (in): the string to search through.<p>
If before = IGNORE,<br>
rep is compared starting at 1. char of text (text[0]),<br>
else starting at 2. char of text (text[1]).
Chars after "after"-delimiter are ignored.
@param rep (in): the string to look for
@param aRepLen (in): the number of bytes in the string to look for
@param before (in): limitation before rep
@param after (in): limitation after rep
@return true, if rep is found and limitation spec is met or rep is empty
*/
bool ItMatchesDelimited(const PRUnichar * aInString, int32_t aInLength,
const PRUnichar * rep, int32_t aRepLen, LIMTYPE before, LIMTYPE after);
/**
@param see ItMatchesDelimited
@return Number of ItMatchesDelimited in text
*/
uint32_t NumberOfMatches(const PRUnichar * aInString, int32_t aInStringLength,
const PRUnichar* rep, int32_t aRepLen, LIMTYPE before, LIMTYPE after);
/**
Currently only changes "<", ">" and "&". All others stay as they are.<p>
"Char" in function name to avoid side effects with nsString(ch)
constructors.
@param ch (in)
@param aStringToAppendto (out) - the string to append the escaped
string to.
@param inAttribute (in) - will escape quotes, too (which is
only needed for attribute values)
*/
void EscapeChar(const PRUnichar ch, nsString& aStringToAppendto,
bool inAttribute);
/**
See EscapeChar. Escapes the string in place.
*/
void EscapeStr(nsString& aInString, bool inAttribute);
/**
Currently only reverts "<", ">" and "&". All others stay as they are.<p>
@param aInString (in) HTML string
@param aStartPos (in) start index into the buffer
@param aLength (in) length of the buffer
@param aOutString (out) unescaped buffer
*/
void UnescapeStr(const PRUnichar * aInString, int32_t aStartPos,
int32_t aLength, nsString& aOutString);
/**
<em>Note</em>: I use different strategies to pass context between the
functions (full text and pos vs. cutted text and col0, glphyTextLen vs.
replaceBefore/-After). It makes some sense, but is hard to understand
(maintain) :-(.
*/
/**
<p><em>Note:</em> replaceBefore + replaceAfter + 1 (for char at pos) chars
in text should be replaced by outputHTML.</p>
<p><em>Note:</em> This function should be able to process a URL on multiple
lines, but currently, ScanForURLs is called for every line, so it can't.</p>
@param text (in): includes possibly a URL
@param pos (in): position in text, where either ":", "." or "@" are found
@param whathasbeendone (in): What the calling ScanTXT did/has to do with the
(not-linkified) text, i.e. usually the "whattodo" parameter.
(Needed to calculate replaceBefore.) NOT what will be done with
the content of the link.
@param outputHTML (out): URL with HTML-a tag
@param replaceBefore (out): Number of chars of URL before pos
@param replaceAfter (out): Number of chars of URL after pos
@return URL found
*/
bool FindURL(const PRUnichar * aInString, int32_t aInLength, const uint32_t pos,
const uint32_t whathasbeendone,
nsString& outputHTML, int32_t& replaceBefore, int32_t& replaceAfter);
enum modetype {
unknown,
RFC1738, /* Check, if RFC1738, APPENDIX compliant,
like "<URL:http://www.mozilla.org>". */
RFC2396E, /* RFC2396, APPENDIX E allows anglebrackets (like
"<http://www.mozilla.org>") (without "URL:") or
quotation marks(like ""http://www.mozilla.org"").
Also allow email addresses without scheme,
e.g. "<mozilla@bucksch.org>" */
freetext, /* assume heading scheme
with "[a-zA-Z][a-zA-Z0-9+\-\.]*:" like "news:"
(see RFC2396, Section 3.1).
Certain characters (see code) or any whitespace
(including linebreaks) end the URL.
Other certain (punctation) characters (see code)
at the end are stripped off. */
abbreviated /* Similar to freetext, but without scheme, e.g.
"www.mozilla.org", "ftp.mozilla.org" and
"mozilla@bucksch.org". */
/* RFC1738 and RFC2396E type URLs may use multiple lines,
whitespace is stripped. Special characters like ")" stay intact.*/
};
/**
* @param text (in), pos (in): see FindURL
* @param check (in): Start must be conform with this mode
* @param start (out): Position in text, where URL (including brackets or
* similar) starts
* @return |check|-conform start has been found
*/
bool FindURLStart(const PRUnichar * aInString, int32_t aInLength, const uint32_t pos,
const modetype check, uint32_t& start);
/**
* @param text (in), pos (in): see FindURL
* @param check (in): End must be conform with this mode
* @param start (in): see FindURLStart
* @param end (out): Similar to |start| param of FindURLStart
* @return |check|-conform end has been found
*/
bool FindURLEnd(const PRUnichar * aInString, int32_t aInStringLength, const uint32_t pos,
const modetype check, const uint32_t start, uint32_t& end);
/**
* @param text (in), pos (in), whathasbeendone (in): see FindURL
* @param check (in): Current mode
* @param start (in), end (in): see FindURLEnd
* @param txtURL (out): Guessed (raw) URL.
* Without whitespace, but not completed.
* @param desc (out): Link as shown to the user, but already escaped.
* Should be placed between the <a> and </a> tags.
* @param replaceBefore(out), replaceAfter (out): see FindURL
*/
void CalculateURLBoundaries(const PRUnichar * aInString, int32_t aInStringLength,
const uint32_t pos, const uint32_t whathasbeendone,
const modetype check, const uint32_t start, const uint32_t end,
nsString& txtURL, nsString& desc,
int32_t& replaceBefore, int32_t& replaceAfter);
/**
* @param txtURL (in), desc (in): see CalculateURLBoundaries
* @param outputHTML (out): see FindURL
* @return A valid URL could be found (and creation of HTML successful)
*/
bool CheckURLAndCreateHTML(
const nsString& txtURL, const nsString& desc, const modetype mode,
nsString& outputHTML);
/**
@param text (in): line of text possibly with tagTXT.<p>
if col0 is true,
starting with tagTXT<br>
else
starting one char before tagTXT
@param col0 (in): tagTXT is on the beginning of the line (or paragraph).
open must be 0 then.
@param tagTXT (in): Tag in plaintext to search for, e.g. "*"
@param aTagTxtLen (in): length of tagTXT.
@param tagHTML (in): HTML-Tag to replace tagTXT with,
without "<" and ">", e.g. "strong"
@param attributeHTML (in): HTML-attribute to add to opening tagHTML,
e.g. "class=txt_star"
@param aOutString: string to APPEND the converted html into
@param open (in/out): Number of currently open tags of type tagHTML
@return Conversion succeeded
*/
bool StructPhraseHit(const PRUnichar * aInString, int32_t aInStringLength, bool col0,
const PRUnichar* tagTXT,
int32_t aTagTxtLen,
const char* tagHTML, const char* attributeHTML,
nsString& aOutputString, uint32_t& openTags);
/**
@param text (in), col0 (in): see GlyphHit
@param tagTXT (in): Smily, see also StructPhraseHit
@param imageName (in): the basename of the file that contains the image for this smilie
@param outputHTML (out): new string containing the html for the smily
@param glyphTextLen (out): see GlyphHit
*/
bool
SmilyHit(const PRUnichar * aInString, int32_t aLength, bool col0,
const char* tagTXT, const char* imageName,
nsString& outputHTML, int32_t& glyphTextLen);
/**
Checks, if we can replace some chars at the start of line with prettier HTML
code.<p>
If success is reported, replace the first glyphTextLen chars with outputHTML
@param text (in): line of text possibly with Glyph.<p>
If col0 is true,
starting with Glyph <br><!-- (br not part of text) -->
else
starting one char before Glyph
@param col0 (in): text starts at the beginning of the line (or paragraph)
@param aOutString (out): APPENDS html for the glyph to this string
@param glyphTextLen (out): Length of original text to replace
@return see StructPhraseHit
*/
bool GlyphHit(const PRUnichar * aInString, int32_t aInLength, bool col0,
nsString& aOutString, int32_t& glyphTextLen);
/**
Check if a given url should be linkified.
@param aURL (in): url to be checked on.
*/
bool ShouldLinkify(const nsCString& aURL);
};
// It's said, that Win32 and Mac don't like static const members
const int32_t mozTXTToHTMLConv_lastMode = 4;
// Needed (only) by mozTXTToHTMLConv::FindURL
const int32_t mozTXTToHTMLConv_numberOfModes = 4; // dito; unknown not counted
#endif