cpl_string.cpp

/**********************************************************************
 *
 * Name:     cpl_string.cpp
 * Project:  CPL - Common Portability Library
 * Purpose:  String and Stringlist manipulation functions.
 * Author:   Daniel Morissette, danmo@videotron.ca
 *
 **********************************************************************
 * Copyright (c) 1998, Daniel Morissette
 * Copyright (c) 2008-2013, Even Rouault <even dot rouault at spatialys.com>
 *
 * SPDX-License-Identifier: MIT
 **********************************************************************
 *
 * Independent Security Audit 2003/04/04 Andrey Kiselev:
 *   Completed audit of this module. All functions may be used without buffer
 *   overflows and stack corruptions with any kind of input data strings with
 *   except of CPLSPrintf() and CSLAppendPrintf() (see note below).
 *
 * Security Audit 2003/03/28 warmerda:
 *   Completed security audit.  I believe that this module may be safely used
 *   to parse tokenize arbitrary input strings, assemble arbitrary sets of
 *   names values into string lists, unescape and escape text even if provided
 *   by a potentially hostile source.
 *
 *   CPLSPrintf() and CSLAppendPrintf() may not be safely invoked on
 *   arbitrary length inputs since it has a fixed size output buffer on system
 *   without vsnprintf().
 *
 **********************************************************************/

#undef WARN_STANDARD_PRINTF

#include "cpl_port.h"
#include "cpl_string.h"

#include <cctype>
#include <climits>
#include <cmath>
#include <cstdlib>
#include <cstring>

#include <limits>

#include "cpl_config.h"
#include "cpl_multiproc.h"
#include "cpl_vsi.h"

#if !defined(va_copy) && defined(__va_copy)
#define va_copy __va_copy
#endif

/*=====================================================================
                    StringList manipulation functions.
 =====================================================================*/

/**********************************************************************
 *                       CSLAddString()
 **********************************************************************/

/** Append a string to a StringList and return a pointer to the modified
 * StringList.
 *
 * If the input StringList is NULL, then a new StringList is created.
 * Note that CSLAddString performance when building a list is in O(n^2)
 * which can cause noticeable slow down when n > 10000.
 */
char **CSLAddString(char **papszStrList, const char *pszNewString)
{
    char **papszRet = CSLAddStringMayFail(papszStrList, pszNewString);
    if (papszRet == nullptr && pszNewString != nullptr)
        abort();
    return papszRet;
}

/** Same as CSLAddString() but may return NULL in case of (memory) failure */
char **CSLAddStringMayFail(char **papszStrList, const char *pszNewString)
{
    if (pszNewString == nullptr)
        return papszStrList;  // Nothing to do!

    char *pszDup = VSI_STRDUP_VERBOSE(pszNewString);
    if (pszDup == nullptr)
        return nullptr;

    // Allocate room for the new string.
    char **papszStrListNew = nullptr;
    int nItems = 0;

    if (papszStrList == nullptr)
        papszStrListNew =
            static_cast<char **>(VSI_CALLOC_VERBOSE(2, sizeof(char *)));
    else
    {
        nItems = CSLCount(papszStrList);
        papszStrListNew = static_cast<char **>(
            VSI_REALLOC_VERBOSE(papszStrList, (nItems + 2) * sizeof(char *)));
    }
    if (papszStrListNew == nullptr)
    {
        VSIFree(pszDup);
        return nullptr;
    }

    // Copy the string in the list.
    papszStrListNew[nItems] = pszDup;
    papszStrListNew[nItems + 1] = nullptr;

    return papszStrListNew;
}

/************************************************************************/
/*                              CSLCount()                              */
/************************************************************************/

/**
 * Return number of items in a string list.
 *
 * Returns the number of items in a string list, not counting the
 * terminating NULL.  Passing in NULL is safe, and will result in a count
 * of zero.
 *
 * Lists are counted by iterating through them so long lists will
 * take more time than short lists.  Care should be taken to avoid using
 * CSLCount() as an end condition for loops as it will result in O(n^2)
 * behavior.
 *
 * @param papszStrList the string list to count.
 *
 * @return the number of entries.
 */
int CSLCount(CSLConstList papszStrList)
{
    if (!papszStrList)
        return 0;

    int nItems = 0;

    while (*papszStrList != nullptr)
    {
        ++nItems;
        ++papszStrList;
    }

    return nItems;
}

/************************************************************************/
/*                            CSLGetField()                             */
/************************************************************************/

/**
 * Fetches the indicated field, being careful not to crash if the field
 * doesn't exist within this string list.
 *
 * The returned pointer should not be freed, and doesn't necessarily last long.
 */
const char *CSLGetField(CSLConstList papszStrList, int iField)

{
    if (papszStrList == nullptr || iField < 0)
        return ("");

    for (int i = 0; i < iField + 1; i++)
    {
        if (papszStrList[i] == nullptr)
            return "";
    }

    return (papszStrList[iField]);
}

/************************************************************************/
/*                             CSLDestroy()                             */
/************************************************************************/

/**
 * Free string list.
 *
 * Frees the passed string list (null terminated array of strings).
 * It is safe to pass NULL.
 *
 * @param papszStrList the list to free.
 */
void CPL_STDCALL CSLDestroy(char **papszStrList)
{
    if (!papszStrList)
        return;

    for (char **papszPtr = papszStrList; *papszPtr != nullptr; ++papszPtr)
    {
        CPLFree(*papszPtr);
    }

    CPLFree(papszStrList);
}

/************************************************************************/
/*                            CSLDuplicate()                            */
/************************************************************************/

/**
 * Clone a string list.
 *
 * Efficiently allocates a copy of a string list.  The returned list is
 * owned by the caller and should be freed with CSLDestroy().
 *
 * @param papszStrList the input string list.
 *
 * @return newly allocated copy.
 */

char **CSLDuplicate(CSLConstList papszStrList)
{
    const int nLines = CSLCount(papszStrList);

    if (nLines == 0)
        return nullptr;

    CSLConstList papszSrc = papszStrList;

    char **papszNewList =
        static_cast<char **>(VSI_MALLOC2_VERBOSE(nLines + 1, sizeof(char *)));

    char **papszDst = papszNewList;

    for (; *papszSrc != nullptr; ++papszSrc, ++papszDst)
    {
        *papszDst = VSI_STRDUP_VERBOSE(*papszSrc);
        if (*papszDst == nullptr)
        {
            CSLDestroy(papszNewList);
            return nullptr;
        }
    }
    *papszDst = nullptr;

    return papszNewList;
}

/************************************************************************/
/*                               CSLMerge                               */
/************************************************************************/

/**
 * \brief Merge two lists.
 *
 * The two lists are merged, ensuring that if any keys appear in both
 * that the value from the second (papszOverride) list take precedence.
 *
 * @param papszOrig the original list, being modified.
 * @param papszOverride the list of items being merged in.  This list
 * is unaltered and remains owned by the caller.
 *
 * @return updated list.
 */

char **CSLMerge(char **papszOrig, CSLConstList papszOverride)

{
    if (papszOrig == nullptr && papszOverride != nullptr)
        return CSLDuplicate(papszOverride);

    if (papszOverride == nullptr)
        return papszOrig;

    for (int i = 0; papszOverride[i] != nullptr; ++i)
    {
        char *pszKey = nullptr;
        const char *pszValue = CPLParseNameValue(papszOverride[i], &pszKey);

        papszOrig = CSLSetNameValue(papszOrig, pszKey, pszValue);
        CPLFree(pszKey);
    }

    return papszOrig;
}

/************************************************************************/
/*                             CSLLoad2()                               */
/************************************************************************/

/**
 * Load a text file into a string list.
 *
 * The VSI*L API is used, so VSIFOpenL() supported objects that aren't
 * physical files can also be accessed.  Files are returned as a string list,
 * with one item in the string list per line.  End of line markers are
 * stripped (by CPLReadLineL()).
 *
 * If reading the file fails a CPLError() will be issued and NULL returned.
 *
 * @param pszFname the name of the file to read.
 * @param nMaxLines maximum number of lines to read before stopping, or -1 for
 * no limit.
 * @param nMaxCols maximum number of characters in a line before stopping, or -1
 * for no limit.
 * @param papszOptions NULL-terminated array of options. Unused for now.
 *
 * @return a string list with the files lines, now owned by caller. To be freed
 * with CSLDestroy()
 *
 * @since GDAL 1.7.0
 */

char **CSLLoad2(const char *pszFname, int nMaxLines, int nMaxCols,
                CSLConstList papszOptions)
{
    VSILFILE *fp = VSIFOpenL(pszFname, "rb");

    if (!fp)
    {
        if (CPLFetchBool(papszOptions, "EMIT_ERROR_IF_CANNOT_OPEN_FILE", true))
        {
            // Unable to open file.
            CPLError(CE_Failure, CPLE_OpenFailed,
                     "CSLLoad2(\"%s\") failed: unable to open file.", pszFname);
        }
        return nullptr;
    }

    char **papszStrList = nullptr;
    int nLines = 0;
    int nAllocatedLines = 0;

    while (!VSIFEofL(fp) && (nMaxLines == -1 || nLines < nMaxLines))
    {
        const char *pszLine = CPLReadLine2L(fp, nMaxCols, papszOptions);
        if (pszLine == nullptr)
            break;

        if (nLines + 1 >= nAllocatedLines)
        {
            nAllocatedLines = 16 + nAllocatedLines * 2;
            char **papszStrListNew = static_cast<char **>(
                VSIRealloc(papszStrList, nAllocatedLines * sizeof(char *)));
            if (papszStrListNew == nullptr)
            {
                CPL_IGNORE_RET_VAL(VSIFCloseL(fp));
                CPLReadLineL(nullptr);
                CPLError(CE_Failure, CPLE_OutOfMemory,
                         "CSLLoad2(\"%s\") "
                         "failed: not enough memory to allocate lines.",
                         pszFname);
                return papszStrList;
            }
            papszStrList = papszStrListNew;
        }
        papszStrList[nLines] = CPLStrdup(pszLine);
        papszStrList[nLines + 1] = nullptr;
        ++nLines;
    }

    CPL_IGNORE_RET_VAL(VSIFCloseL(fp));

    // Free the internal thread local line buffer.
    CPLReadLineL(nullptr);

    return papszStrList;
}

/************************************************************************/
/*                              CSLLoad()                               */
/************************************************************************/

/**
 * Load a text file into a string list.
 *
 * The VSI*L API is used, so VSIFOpenL() supported objects that aren't
 * physical files can also be accessed.  Files are returned as a string list,
 * with one item in the string list per line.  End of line markers are
 * stripped (by CPLReadLineL()).
 *
 * If reading the file fails a CPLError() will be issued and NULL returned.
 *
 * @param pszFname the name of the file to read.
 *
 * @return a string list with the files lines, now owned by caller. To be freed
 * with CSLDestroy()
 */

char **CSLLoad(const char *pszFname)
{
    return CSLLoad2(pszFname, -1, -1, nullptr);
}

/**********************************************************************
 *                       CSLSave()
 **********************************************************************/

/** Write a StringList to a text file.
 *
 * Returns the number of lines written, or 0 if the file could not
 * be written.
 */

int CSLSave(CSLConstList papszStrList, const char *pszFname)
{
    if (papszStrList == nullptr)
        return 0;

    VSILFILE *fp = VSIFOpenL(pszFname, "wt");
    if (fp == nullptr)
    {
        // Unable to open file.
        CPLError(CE_Failure, CPLE_OpenFailed,
                 "CSLSave(\"%s\") failed: unable to open output file.",
                 pszFname);
        return 0;
    }

    int nLines = 0;
    while (*papszStrList != nullptr)
    {
        if (VSIFPrintfL(fp, "%s\n", *papszStrList) < 1)
        {
            CPLError(CE_Failure, CPLE_FileIO,
                     "CSLSave(\"%s\") failed: unable to write to output file.",
                     pszFname);
            break;  // A Problem happened... abort.
        }

        ++nLines;
        ++papszStrList;
    }

    if (VSIFCloseL(fp) != 0)
    {
        CPLError(CE_Failure, CPLE_FileIO,
                 "CSLSave(\"%s\") failed: unable to write to output file.",
                 pszFname);
    }

    return nLines;
}

/**********************************************************************
 *                       CSLPrint()
 **********************************************************************/

/** Print a StringList to fpOut.  If fpOut==NULL, then output is sent
 * to stdout.
 *
 * Returns the number of lines printed.
 */
int CSLPrint(CSLConstList papszStrList, FILE *fpOut)
{
    if (!papszStrList)
        return 0;

    if (fpOut == nullptr)
        fpOut = stdout;

    int nLines = 0;

    while (*papszStrList != nullptr)
    {
        if (VSIFPrintf(fpOut, "%s\n", *papszStrList) < 0)
            return nLines;
        ++nLines;
        ++papszStrList;
    }

    return nLines;
}

/**********************************************************************
 *                       CSLInsertStrings()
 **********************************************************************/

/** Copies the contents of a StringList inside another StringList
 * before the specified line.
 *
 * nInsertAtLineNo is a 0-based line index before which the new strings
 * should be inserted.  If this value is -1 or is larger than the actual
 * number of strings in the list then the strings are added at the end
 * of the source StringList.
 *
 * Returns the modified StringList.
 */

char **CSLInsertStrings(char **papszStrList, int nInsertAtLineNo,
                        CSLConstList papszNewLines)
{
    if (papszNewLines == nullptr)
        return papszStrList;  // Nothing to do!

    const int nToInsert = CSLCount(papszNewLines);
    if (nToInsert == 0)
        return papszStrList;  // Nothing to do!

    const int nSrcLines = CSLCount(papszStrList);
    const int nDstLines = nSrcLines + nToInsert;

    // Allocate room for the new strings.
    papszStrList = static_cast<char **>(
        CPLRealloc(papszStrList, (nDstLines + 1) * sizeof(char *)));

    // Make sure the array is NULL-terminated.  It may not be if
    // papszStrList was NULL before Realloc().
    papszStrList[nSrcLines] = nullptr;

    // Make some room in the original list at the specified location.
    // Note that we also have to move the NULL pointer at the end of
    // the source StringList.
    if (nInsertAtLineNo == -1 || nInsertAtLineNo > nSrcLines)
        nInsertAtLineNo = nSrcLines;

    {
        char **ppszSrc = papszStrList + nSrcLines;
        char **ppszDst = papszStrList + nDstLines;

        for (int i = nSrcLines; i >= nInsertAtLineNo; --i)
        {
            *ppszDst = *ppszSrc;
            --ppszDst;
            --ppszSrc;
        }
    }

    // Copy the strings to the list.
    CSLConstList ppszSrc = papszNewLines;
    char **ppszDst = papszStrList + nInsertAtLineNo;

    for (; *ppszSrc != nullptr; ++ppszSrc, ++ppszDst)
    {
        *ppszDst = CPLStrdup(*ppszSrc);
    }

    return papszStrList;
}

/**********************************************************************
 *                       CSLInsertString()
 **********************************************************************/

/** Insert a string at a given line number inside a StringList
 *
 * nInsertAtLineNo is a 0-based line index before which the new string
 * should be inserted.  If this value is -1 or is larger than the actual
 * number of strings in the list then the string is added at the end
 * of the source StringList.
 *
 * Returns the modified StringList.
 */

char **CSLInsertString(char **papszStrList, int nInsertAtLineNo,
                       const char *pszNewLine)
{
    char *apszList[2] = {const_cast<char *>(pszNewLine), nullptr};

    return CSLInsertStrings(papszStrList, nInsertAtLineNo, apszList);
}

/**********************************************************************
 *                       CSLRemoveStrings()
 **********************************************************************/

/** Remove strings inside a StringList
 *
 * nFirstLineToDelete is the 0-based line index of the first line to
 * remove. If this value is -1 or is larger than the actual
 * number of strings in list then the nNumToRemove last strings are
 * removed.
 *
 * If ppapszRetStrings != NULL then the deleted strings won't be
 * free'd, they will be stored in a new StringList and the pointer to
 * this new list will be returned in *ppapszRetStrings.
 *
 * Returns the modified StringList.
 */

char **CSLRemoveStrings(char **papszStrList, int nFirstLineToDelete,
                        int nNumToRemove, char ***ppapszRetStrings)
{
    const int nSrcLines = CSLCount(papszStrList);

    if (nNumToRemove < 1 || nSrcLines == 0)
        return papszStrList;  // Nothing to do!

    // If operation will result in an empty StringList, don't waste
    // time here.
    const int nDstLines = nSrcLines - nNumToRemove;
    if (nDstLines < 1)
    {
        CSLDestroy(papszStrList);
        return nullptr;
    }

    // Remove lines from the source StringList.
    // Either free() each line or store them to a new StringList depending on
    // the caller's choice.
    char **ppszDst = papszStrList + nFirstLineToDelete;

    if (ppapszRetStrings == nullptr)
    {
        // free() all the strings that will be removed.
        for (int i = 0; i < nNumToRemove; ++i)
        {
            CPLFree(*ppszDst);
            *ppszDst = nullptr;
        }
    }
    else
    {
        // Store the strings to remove in a new StringList.
        *ppapszRetStrings =
            static_cast<char **>(CPLCalloc(nNumToRemove + 1, sizeof(char *)));

        for (int i = 0; i < nNumToRemove; ++i)
        {
            (*ppapszRetStrings)[i] = *ppszDst;
            *ppszDst = nullptr;
            ++ppszDst;
        }
    }

    // Shift down all the lines that follow the lines to remove.
    if (nFirstLineToDelete == -1 || nFirstLineToDelete > nSrcLines)
        nFirstLineToDelete = nDstLines;

    char **ppszSrc = papszStrList + nFirstLineToDelete + nNumToRemove;
    ppszDst = papszStrList + nFirstLineToDelete;

    for (; *ppszSrc != nullptr; ++ppszSrc, ++ppszDst)
    {
        *ppszDst = *ppszSrc;
    }
    // Move the NULL pointer at the end of the StringList.
    *ppszDst = *ppszSrc;

    // At this point, we could realloc() papszStrList to a smaller size, but
    // since this array will likely grow again in further operations on the
    // StringList we'll leave it as it is.
    return papszStrList;
}

/************************************************************************/
/*                           CSLFindString()                            */
/************************************************************************/

/**
 * Find a string within a string list (case insensitive).
 *
 * Returns the index of the entry in the string list that contains the
 * target string.  The string in the string list must be a full match for
 * the target, but the search is case insensitive.
 *
 * @param papszList the string list to be searched.
 * @param pszTarget the string to be searched for.
 *
 * @return the index of the string within the list or -1 on failure.
 */

int CSLFindString(CSLConstList papszList, const char *pszTarget)

{
    if (papszList == nullptr)
        return -1;

    for (int i = 0; papszList[i] != nullptr; ++i)
    {
        if (EQUAL(papszList[i], pszTarget))
            return i;
    }

    return -1;
}

/************************************************************************/
/*                     CSLFindStringCaseSensitive()                     */
/************************************************************************/

/**
 * Find a string within a string list(case sensitive)
 *
 * Returns the index of the entry in the string list that contains the
 * target string.  The string in the string list must be a full match for
 * the target.
 *
 * @param papszList the string list to be searched.
 * @param pszTarget the string to be searched for.
 *
 * @return the index of the string within the list or -1 on failure.
 *
 * @since GDAL 2.0
 */

int CSLFindStringCaseSensitive(CSLConstList papszList, const char *pszTarget)

{
    if (papszList == nullptr)
        return -1;

    for (int i = 0; papszList[i] != nullptr; ++i)
    {
        if (strcmp(papszList[i], pszTarget) == 0)
            return i;
    }

    return -1;
}

/************************************************************************/
/*                           CSLPartialFindString()                     */
/************************************************************************/

/**
 * Find a substring within a string list.
 *
 * Returns the index of the entry in the string list that contains the
 * target string as a substring.  The search is case sensitive (unlike
 * CSLFindString()).
 *
 * @param papszHaystack the string list to be searched.
 * @param pszNeedle the substring to be searched for.
 *
 * @return the index of the string within the list or -1 on failure.
 */

int CSLPartialFindString(CSLConstList papszHaystack, const char *pszNeedle)
{
    if (papszHaystack == nullptr || pszNeedle == nullptr)
        return -1;

    for (int i = 0; papszHaystack[i] != nullptr; ++i)
    {
        if (strstr(papszHaystack[i], pszNeedle))
            return i;
    }

    return -1;
}

/**********************************************************************
 *                       CSLTokenizeString()
 **********************************************************************/

/** Tokenizes a string and returns a StringList with one string for
 * each token.
 */
char **CSLTokenizeString(const char *pszString)
{
    return CSLTokenizeString2(pszString, " ", CSLT_HONOURSTRINGS);
}

/************************************************************************/
/*                      CSLTokenizeStringComplex()                      */
/************************************************************************/

/** Obsolete tokenizing api. Use CSLTokenizeString2() */
char **CSLTokenizeStringComplex(const char *pszString,
                                const char *pszDelimiters, int bHonourStrings,
                                int bAllowEmptyTokens)
{
    int nFlags = 0;

    if (bHonourStrings)
        nFlags |= CSLT_HONOURSTRINGS;
    if (bAllowEmptyTokens)
        nFlags |= CSLT_ALLOWEMPTYTOKENS;

    return CSLTokenizeString2(pszString, pszDelimiters, nFlags);
}

/************************************************************************/
/*                         CSLTokenizeString2()                         */
/************************************************************************/

/**
 * Tokenize a string.
 *
 * This function will split a string into tokens based on specified'
 * delimiter(s) with a variety of options.  The returned result is a
 * string list that should be freed with CSLDestroy() when no longer
 * needed.
 *
 * The available parsing options are:
 *
 * - CSLT_ALLOWEMPTYTOKENS: allow the return of empty tokens when two
 * delimiters in a row occur with no other text between them.  If not set,
 * empty tokens will be discarded;
 * - CSLT_STRIPLEADSPACES: strip leading space characters from the token (as
 * reported by isspace());
 * - CSLT_STRIPENDSPACES: strip ending space characters from the token (as
 * reported by isspace());
 * - CSLT_HONOURSTRINGS: double quotes can be used to hold values that should
 * not be broken into multiple tokens;
 * - CSLT_PRESERVEQUOTES: string quotes are carried into the tokens when this
 * is set, otherwise they are removed;
 * - CSLT_PRESERVEESCAPES: if set backslash escapes (for backslash itself,
 * and for literal double quotes) will be preserved in the tokens, otherwise
 * the backslashes will be removed in processing.
 *
 * \b Example:
 *
 * Parse a string into tokens based on various white space (space, newline,
 * tab) and then print out results and cleanup.  Quotes may be used to hold
 * white space in tokens.

\code
    char **papszTokens =
        CSLTokenizeString2( pszCommand, " \t\n",
                            CSLT_HONOURSTRINGS | CSLT_ALLOWEMPTYTOKENS );

    for( int i = 0; papszTokens != NULL && papszTokens[i] != NULL; ++i )
        printf( "arg %d: '%s'", papszTokens[i] );  // ok

    CSLDestroy( papszTokens );
\endcode

 * @param pszString the string to be split into tokens.
 * @param pszDelimiters one or more characters to be used as token delimiters.
 * @param nCSLTFlags an ORing of one or more of the CSLT_ flag values.
 *
 * @return a string list of tokens owned by the caller.
 */

char **CSLTokenizeString2(const char *pszString, const char *pszDelimiters,
                          int nCSLTFlags)
{
    if (pszString == nullptr)
        return static_cast<char **>(CPLCalloc(sizeof(char *), 1));

    CPLStringList oRetList;
    const bool bHonourStrings = (nCSLTFlags & CSLT_HONOURSTRINGS) != 0;
    const bool bAllowEmptyTokens = (nCSLTFlags & CSLT_ALLOWEMPTYTOKENS) != 0;
    const bool bStripLeadSpaces = (nCSLTFlags & CSLT_STRIPLEADSPACES) != 0;
    const bool bStripEndSpaces = (nCSLTFlags & CSLT_STRIPENDSPACES) != 0;

    char *pszToken = static_cast<char *>(CPLCalloc(10, 1));
    size_t nTokenMax = 10;

    while (*pszString != '\0')
    {
        bool bInString = false;
        bool bStartString = true;
        size_t nTokenLen = 0;

        // Try to find the next delimiter, marking end of token.
        for (; *pszString != '\0'; ++pszString)
        {
            // Extend token buffer if we are running close to its end.
            if (nTokenLen >= nTokenMax - 3)
            {
                if (nTokenMax > std::numeric_limits<size_t>::max() / 2)
                {
                    CPLFree(pszToken);
                    return static_cast<char **>(CPLCalloc(sizeof(char *), 1));
                }
                nTokenMax = nTokenMax * 2;
                char *pszNewToken = static_cast<char *>(
                    VSI_REALLOC_VERBOSE(pszToken, nTokenMax));
                if (pszNewToken == nullptr)
                {
                    CPLFree(pszToken);
                    return static_cast<char **>(CPLCalloc(sizeof(char *), 1));
                }
                pszToken = pszNewToken;
            }

            // End if this is a delimiter skip it and break.
            if (!bInString && strchr(pszDelimiters, *pszString) != nullptr)
            {
                ++pszString;
                break;
            }

            // If this is a quote, and we are honouring constant
            // strings, then process the constant strings, with out delim
            // but don't copy over the quotes.
            if (bHonourStrings && *pszString == '"')
            {
                if (nCSLTFlags & CSLT_PRESERVEQUOTES)
                {
                    pszToken[nTokenLen] = *pszString;
                    ++nTokenLen;
                }

                bInString = !bInString;
                continue;
            }

            /*
             * Within string constants we allow for escaped quotes, but in
             * processing them we will unescape the quotes and \\ sequence
             * reduces to \
             */
            if (bInString && pszString[0] == '\\')
            {
                if (pszString[1] == '"' || pszString[1] == '\\')
                {
                    if (nCSLTFlags & CSLT_PRESERVEESCAPES)
                    {
                        pszToken[nTokenLen] = *pszString;
                        ++nTokenLen;
                    }

                    ++pszString;
                }
            }

            // Strip spaces at the token start if requested.
            if (!bInString && bStripLeadSpaces && bStartString &&
                isspace(static_cast<unsigned char>(*pszString)))
                continue;

            bStartString = false;

            pszToken[nTokenLen] = *pszString;
            ++nTokenLen;
        }

        // Strip spaces at the token end if requested.
        if (!bInString && bStripEndSpaces)
        {
            while (nTokenLen &&
                   isspace(static_cast<unsigned char>(pszToken[nTokenLen - 1])))
                nTokenLen--;
        }

        pszToken[nTokenLen] = '\0';

        // Add the token.
        if (pszToken[0] != '\0' || bAllowEmptyTokens)
            oRetList.AddString(pszToken);
    }

    /*
     * If the last token was empty, then we need to capture
     * it now, as the loop would skip it.
     */
    if (*pszString == '\0' && bAllowEmptyTokens && oRetList.Count() > 0 &&
        strchr(pszDelimiters, *(pszString - 1)) != nullptr)
    {
        oRetList.AddString("");
    }

    CPLFree(pszToken);

    if (oRetList.List() == nullptr)
    {
        // Prefer to return empty lists as a pointer to
        // a null pointer since some client code might depend on this.
        oRetList.Assign(static_cast<char **>(CPLCalloc(sizeof(char *), 1)));
    }

    return oRetList.StealList();
}

/**********************************************************************
 *                       CPLSPrintf()
 *
 * NOTE: This function should move to cpl_conv.cpp.
 **********************************************************************/

// For now, assume that a 8000 chars buffer will be enough.
constexpr int CPLSPrintf_BUF_SIZE = 8000;
constexpr int CPLSPrintf_BUF_Count = 10;

/** CPLSPrintf() that works with 10 static buffer.
 *
 * It returns a ref. to a static buffer that should not be freed and
 * is valid only until the next call to CPLSPrintf().
 */

const char *CPLSPrintf(CPL_FORMAT_STRING(const char *fmt), ...)
{
    va_list args;

    /* -------------------------------------------------------------------- */
    /*      Get the thread local buffer ring data.                          */
    /* -------------------------------------------------------------------- */
    char *pachBufRingInfo = static_cast<char *>(CPLGetTLS(CTLS_CPLSPRINTF));

    if (pachBufRingInfo == nullptr)
    {
        pachBufRingInfo = static_cast<char *>(CPLCalloc(
            1, sizeof(int) + CPLSPrintf_BUF_Count * CPLSPrintf_BUF_SIZE));
        CPLSetTLS(CTLS_CPLSPRINTF, pachBufRingInfo, TRUE);
    }

    /* -------------------------------------------------------------------- */
    /*      Work out which string in the "ring" we want to use this         */
    /*      time.                                                           */
    /* -------------------------------------------------------------------- */
    int *pnBufIndex = reinterpret_cast<int *>(pachBufRingInfo);
    const size_t nOffset = sizeof(int) + *pnBufIndex * CPLSPrintf_BUF_SIZE;
    char *pachBuffer = pachBufRingInfo + nOffset;

    *pnBufIndex = (*pnBufIndex + 1) % CPLSPrintf_BUF_Count;

    /* -------------------------------------------------------------------- */
    /*      Format the result.                                              */
    /* -------------------------------------------------------------------- */

    va_start(args, fmt);

    const int ret =
        CPLvsnprintf(pachBuffer, CPLSPrintf_BUF_SIZE - 1, fmt, args);
    if (ret < 0 || ret >= CPLSPrintf_BUF_SIZE - 1)
    {
        CPLError(CE_Failure, CPLE_AppDefined,
                 "CPLSPrintf() called with too "
                 "big string. Output will be truncated !");
    }

    va_end(args);

    return pachBuffer;
}

/**********************************************************************
 *                       CSLAppendPrintf()
 **********************************************************************/

/** Use CPLSPrintf() to append a new line at the end of a StringList.
 * Returns the modified StringList.
 */
char **CSLAppendPrintf(char **papszStrList, CPL_FORMAT_STRING(const char *fmt),
                       ...)
{
    va_list args;

    va_start(args, fmt);
    CPLString osWork;
    osWork.vPrintf(fmt, args);
    va_end(args);

    return CSLAddString(papszStrList, osWork);
}

/************************************************************************/
/*                            CPLVASPrintf()                            */
/************************************************************************/

/** This is intended to serve as an easy to use C callable vasprintf()
 * alternative.  Used in the GeoJSON library for instance */
int CPLVASPrintf(char **buf, CPL_FORMAT_STRING(const char *fmt), va_list ap)

{
    CPLString osWork;

    osWork.vPrintf(fmt, ap);

    if (buf)
        *buf = CPLStrdup(osWork.c_str());

    return static_cast<int>(osWork.size());
}

/************************************************************************/
/*                  CPLvsnprintf_get_end_of_formatting()                */
/************************************************************************/

static const char *CPLvsnprintf_get_end_of_formatting(const char *fmt)
{
    char ch = '\0';
    // Flag.
    for (; (ch = *fmt) != '\0'; ++fmt)
    {
        if (ch == '\'')
            continue;  // Bad idea as this is locale specific.
        if (ch == '-' || ch == '+' || ch == ' ' || ch == '#' || ch == '0')
            continue;
        break;
    }

    // Field width.
    for (; (ch = *fmt) != '\0'; ++fmt)
    {
        if (ch == '$')
            return nullptr;  // Do not support this.
        if (*fmt >= '0' && *fmt <= '9')
            continue;
        break;
    }

    // Precision.
    if (ch == '.')
    {
        ++fmt;
        for (; (ch = *fmt) != '\0'; ++fmt)
        {
            if (ch == '$')
                return nullptr;  // Do not support this.
            if (*fmt >= '0' && *fmt <= '9')
                continue;
            break;
        }
    }

    // Length modifier.
    for (; (ch = *fmt) != '\0'; ++fmt)
    {
        if (ch == 'h' || ch == 'l' || ch == 'j' || ch == 'z' || ch == 't' ||
            ch == 'L')
            continue;
        else if (ch == 'I' && fmt[1] == '6' && fmt[2] == '4')
            fmt += 2;
        else
            return fmt;
    }

    return nullptr;
}

/************************************************************************/
/*                           CPLvsnprintf()                             */
/************************************************************************/

#define call_native_snprintf(type)                                             \
    local_ret = snprintf(str + offset_out, size - offset_out, localfmt,        \
                         va_arg(wrk_args, type))

/** vsnprintf() wrapper that is not sensitive to LC_NUMERIC settings.
 *
 * This function has the same contract as standard vsnprintf(), except that
 * formatting of floating-point numbers will use decimal point, whatever the
 * current locale is set.
 *
 * @param str output buffer
 * @param size size of the output buffer (including space for terminating nul)
 * @param fmt formatting string
 * @param args arguments
 * @return the number of characters (excluding terminating nul) that would be
 * written if size is big enough. Or potentially -1 with Microsoft C runtime
 * for Visual Studio < 2015.
 * @since GDAL 2.0
 */
int CPLvsnprintf(char *str, size_t size, CPL_FORMAT_STRING(const char *fmt),
                 va_list args)
{
    if (size == 0)
        return vsnprintf(str, size, fmt, args);

    va_list wrk_args;

#ifdef va_copy
    va_copy(wrk_args, args);
#else
    wrk_args = args;
#endif

    const char *fmt_ori = fmt;
    size_t offset_out = 0;
    char ch = '\0';
    bool bFormatUnknown = false;

    for (; (ch = *fmt) != '\0'; ++fmt)
    {
        if (ch == '%')
        {
            if (strncmp(fmt, "%.*f", 4) == 0)
            {
                const int precision = va_arg(wrk_args, int);
                const double val = va_arg(wrk_args, double);
                const int local_ret =
                    snprintf(str + offset_out, size - offset_out, "%.*f",
                             precision, val);
                // MSVC vsnprintf() returns -1.
                if (local_ret < 0 || offset_out + local_ret >= size)
                    break;
                for (int j = 0; j < local_ret; ++j)
                {
                    if (str[offset_out + j] == ',')
                    {
                        str[offset_out + j] = '.';
                        break;
                    }
                }
                offset_out += local_ret;
                fmt += strlen("%.*f") - 1;
                continue;
            }

            const char *ptrend = CPLvsnprintf_get_end_of_formatting(fmt + 1);
            if (ptrend == nullptr || ptrend - fmt >= 20)
            {
                bFormatUnknown = true;
                break;
            }
            char end = *ptrend;
            char end_m1 = ptrend[-1];

            char localfmt[22] = {};
            memcpy(localfmt, fmt, ptrend - fmt + 1);
            localfmt[ptrend - fmt + 1] = '\0';

            int local_ret = 0;
            if (end == '%')
            {
                if (offset_out == size - 1)
                    break;
                local_ret = 1;
                str[offset_out] = '%';
            }
            else if (end == 'd' || end == 'i' || end == 'c')
            {
                if (end_m1 == 'h')
                    call_native_snprintf(int);
                else if (end_m1 == 'l' && ptrend[-2] != 'l')
                    call_native_snprintf(long);
                else if (end_m1 == 'l' && ptrend[-2] == 'l')
                    call_native_snprintf(GIntBig);
                else if (end_m1 == '4' && ptrend[-2] == '6' &&
                         ptrend[-3] == 'I')
                    // Microsoft I64 modifier.
                    call_native_snprintf(GIntBig);
                else if (end_m1 == 'z')
                    call_native_snprintf(size_t);
                else if ((end_m1 >= 'a' && end_m1 <= 'z') ||
                         (end_m1 >= 'A' && end_m1 <= 'Z'))
                {
                    bFormatUnknown = true;
                    break;
                }
                else
                    call_native_snprintf(int);
            }
            else if (end == 'o' || end == 'u' || end == 'x' || end == 'X')
            {
                if (end_m1 == 'h')
                    call_native_snprintf(unsigned int);
                else if (end_m1 == 'l' && ptrend[-2] != 'l')
                    call_native_snprintf(unsigned long);
                else if (end_m1 == 'l' && ptrend[-2] == 'l')
                    call_native_snprintf(GUIntBig);
                else if (end_m1 == '4' && ptrend[-2] == '6' &&
                         ptrend[-3] == 'I')
                    // Microsoft I64 modifier.
                    call_native_snprintf(GUIntBig);
                else if (end_m1 == 'z')
                    call_native_snprintf(size_t);
                else if ((end_m1 >= 'a' && end_m1 <= 'z') ||
                         (end_m1 >= 'A' && end_m1 <= 'Z'))
                {
                    bFormatUnknown = true;
                    break;
                }
                else
                    call_native_snprintf(unsigned int);
            }
            else if (end == 'e' || end == 'E' || end == 'f' || end == 'F' ||
                     end == 'g' || end == 'G' || end == 'a' || end == 'A')
            {
                if (end_m1 == 'L')
                    call_native_snprintf(long double);
                else
                    call_native_snprintf(double);
                // MSVC vsnprintf() returns -1.
                if (local_ret < 0 || offset_out + local_ret >= size)
                    break;
                for (int j = 0; j < local_ret; ++j)
                {
                    if (str[offset_out + j] == ',')
                    {
                        str[offset_out + j] = '.';
                        break;
                    }
                }
            }
            else if (end == 's')
            {
                const char *pszPtr = va_arg(wrk_args, const char *);
                CPLAssert(pszPtr);
                local_ret = snprintf(str + offset_out, size - offset_out,
                                     localfmt, pszPtr);
            }
            else if (end == 'p')
            {
                call_native_snprintf(void *);
            }
            else
            {
                bFormatUnknown = true;
                break;
            }
            // MSVC vsnprintf() returns -1.
            if (local_ret < 0 || offset_out + local_ret >= size)
                break;
            offset_out += local_ret;
            fmt = ptrend;
        }
        else
        {
            if (offset_out == size - 1)
                break;
            str[offset_out++] = *fmt;
        }
    }
    if (ch == '\0' && offset_out < size)
        str[offset_out] = '\0';
    else
    {
        if (bFormatUnknown)
        {
            CPLDebug("CPL",
                     "CPLvsnprintf() called with unsupported "
                     "formatting string: %s",
                     fmt_ori);
        }
#ifdef va_copy
        va_end(wrk_args);
        va_copy(wrk_args, args);
#else
        wrk_args = args;
#endif
#if defined(HAVE_VSNPRINTF)
        offset_out = vsnprintf(str, size, fmt_ori, wrk_args);
#else
        offset_out = vsprintf(str, fmt_ori, wrk_args);
#endif
    }

#ifdef va_copy
    va_end(wrk_args);
#endif

    return static_cast<int>(offset_out);
}

/************************************************************************/
/*                           CPLsnprintf()                              */
/************************************************************************/

#if !defined(ALIAS_CPLSNPRINTF_AS_SNPRINTF)

#if defined(__clang__) && __clang_major__ == 3 && __clang_minor__ <= 2
#pragma clang diagnostic push
#pragma clang diagnostic ignored "-Wunknown-pragmas"
#pragma clang diagnostic ignored "-Wdocumentation"
#endif

/** snprintf() wrapper that is not sensitive to LC_NUMERIC settings.
 *
 * This function has the same contract as standard snprintf(), except that
 * formatting of floating-point numbers will use decimal point, whatever the
 * current locale is set.
 *
 * @param str output buffer
 * @param size size of the output buffer (including space for terminating nul)
 * @param fmt formatting string
 * @param ... arguments
 * @return the number of characters (excluding terminating nul) that would be
 * written if size is big enough. Or potentially -1 with Microsoft C runtime
 * for Visual Studio < 2015.
 * @since GDAL 2.0
 */

int CPLsnprintf(char *str, size_t size, CPL_FORMAT_STRING(const char *fmt), ...)
{
    va_list args;

    va_start(args, fmt);
    const int ret = CPLvsnprintf(str, size, fmt, args);
    va_end(args);
    return ret;
}

#endif  //  !defined(ALIAS_CPLSNPRINTF_AS_SNPRINTF)

/************************************************************************/
/*                           CPLsprintf()                               */
/************************************************************************/

/** sprintf() wrapper that is not sensitive to LC_NUMERIC settings.
  *
  * This function has the same contract as standard sprintf(), except that
  * formatting of floating-point numbers will use decimal point, whatever the
  * current locale is set.
  *
  * @param str output buffer (must be large enough to hold the result)
  * @param fmt formatting string
  * @param ... arguments
  * @return the number of characters (excluding terminating nul) written in
` * output buffer.
  * @since GDAL 2.0
  */
int CPLsprintf(char *str, CPL_FORMAT_STRING(const char *fmt), ...)
{
    va_list args;

    va_start(args, fmt);
    const int ret = CPLvsnprintf(str, INT_MAX, fmt, args);
    va_end(args);
    return ret;
}

/************************************************************************/
/*                           CPLprintf()                                */
/************************************************************************/

/** printf() wrapper that is not sensitive to LC_NUMERIC settings.
 *
 * This function has the same contract as standard printf(), except that
 * formatting of floating-point numbers will use decimal point, whatever the
 * current locale is set.
 *
 * @param fmt formatting string
 * @param ... arguments
 * @return the number of characters (excluding terminating nul) written in
 * output buffer.
 * @since GDAL 2.0
 */
int CPLprintf(CPL_FORMAT_STRING(const char *fmt), ...)
{
    va_list wrk_args, args;

    va_start(args, fmt);

#ifdef va_copy
    va_copy(wrk_args, args);
#else
    wrk_args = args;
#endif

    char szBuffer[4096] = {};
    // Quiet coverity by staring off nul terminated.
    int ret = CPLvsnprintf(szBuffer, sizeof(szBuffer), fmt, wrk_args);

#ifdef va_copy
    va_end(wrk_args);
#endif

    if (ret < int(sizeof(szBuffer)) - 1)
        ret = printf("%s", szBuffer); /*ok*/
    else
    {
#ifdef va_copy
        va_copy(wrk_args, args);
#else
        wrk_args = args;
#endif

        ret = vfprintf(stdout, fmt, wrk_args);

#ifdef va_copy
        va_end(wrk_args);
#endif
    }

    va_end(args);

    return ret;
}

/************************************************************************/
/*                           CPLsscanf()                                */
/************************************************************************/

/** \brief sscanf() wrapper that is not sensitive to LC_NUMERIC settings.
 *
 * This function has the same contract as standard sscanf(), except that
 * formatting of floating-point numbers will use decimal point, whatever the
 * current locale is set.
 *
 * CAUTION: only works with a very limited number of formatting strings,
 * consisting only of "%lf" and regular characters.
 *
 * @param str input string
 * @param fmt formatting string
 * @param ... arguments
 * @return the number of matched patterns;
 * @since GDAL 2.0
 */
#ifdef DOXYGEN_XML
int CPLsscanf(const char *str, const char *fmt, ...)
#else
int CPLsscanf(const char *str, CPL_SCANF_FORMAT_STRING(const char *fmt), ...)
#endif
{
    bool error = false;
    int ret = 0;
    const char *fmt_ori = fmt;
    va_list args;

    va_start(args, fmt);
    for (; *fmt != '\0' && *str != '\0'; ++fmt)
    {
        if (*fmt == '%')
        {
            if (fmt[1] == 'l' && fmt[2] == 'f')
            {
                fmt += 2;
                char *end;
                *(va_arg(args, double *)) = CPLStrtod(str, &end);
                if (end > str)
                {
                    ++ret;
                    str = end;
                }
                else
                    break;
            }
            else
            {
                error = true;
                break;
            }
        }
        else if (isspace(static_cast<unsigned char>(*fmt)))
        {
            while (*str != '\0' && isspace(static_cast<unsigned char>(*str)))
                ++str;
        }
        else if (*str != *fmt)
            break;
        else
            ++str;
    }
    va_end(args);

    if (error)
    {
        CPLError(CE_Failure, CPLE_NotSupported,
                 "Format %s not supported by CPLsscanf()", fmt_ori);
    }

    return ret;
}

#if defined(__clang__) && __clang_major__ == 3 && __clang_minor__ <= 2
#pragma clang diagnostic pop
#endif

/************************************************************************/
/*                         CPLTestBool()                                */
/************************************************************************/

/**
 * Test what boolean value contained in the string.
 *
 * If pszValue is "NO", "FALSE", "OFF" or "0" will be returned false.
 * Otherwise, true will be returned.
 *
 * @param pszValue the string should be tested.
 *
 * @return true or false.
 */

bool CPLTestBool(const char *pszValue)
{
    return !(EQUAL(pszValue, "NO") || EQUAL(pszValue, "FALSE") ||
             EQUAL(pszValue, "OFF") || EQUAL(pszValue, "0"));
}

/************************************************************************/
/*                         CSLTestBoolean()                             */
/************************************************************************/

/**
 * Test what boolean value contained in the string.
 *
 * If pszValue is "NO", "FALSE", "OFF" or "0" will be returned FALSE.
 * Otherwise, TRUE will be returned.
 *
 * Deprecated.  Removed in GDAL 3.x.
 *
 * Use CPLTestBoolean() for C and CPLTestBool() for C++.
 *
 * @param pszValue the string should be tested.
 *
 * @return TRUE or FALSE.
 */

int CSLTestBoolean(const char *pszValue)
{
    return CPLTestBool(pszValue) ? TRUE : FALSE;
}

/************************************************************************/
/*                         CPLTestBoolean()                             */
/************************************************************************/

/**
 * Test what boolean value contained in the string.
 *
 * If pszValue is "NO", "FALSE", "OFF" or "0" will be returned FALSE.
 * Otherwise, TRUE will be returned.
 *
 * Use this only in C code.  In C++, prefer CPLTestBool().
 *
 * @param pszValue the string should be tested.
 *
 * @return TRUE or FALSE.
 */

int CPLTestBoolean(const char *pszValue)
{
    return CPLTestBool(pszValue) ? TRUE : FALSE;
}

/**********************************************************************
 *                       CPLFetchBool()
 **********************************************************************/

/** Check for boolean key value.
 *
 * In a StringList of "Name=Value" pairs, look to see if there is a key
 * with the given name, and if it can be interpreted as being TRUE.  If
 * the key appears without any "=Value" portion it will be considered true.
 * If the value is NO, FALSE or 0 it will be considered FALSE otherwise
 * if the key appears in the list it will be considered TRUE.  If the key
 * doesn't appear at all, the indicated default value will be returned.
 *
 * @param papszStrList the string list to search.
 * @param pszKey the key value to look for (case insensitive).
 * @param bDefault the value to return if the key isn't found at all.
 *
 * @return true or false
 */

bool CPLFetchBool(CSLConstList papszStrList, const char *pszKey, bool bDefault)

{
    if (CSLFindString(papszStrList, pszKey) != -1)
        return true;

    const char *const pszValue = CSLFetchNameValue(papszStrList, pszKey);
    if (pszValue == nullptr)
        return bDefault;

    return CPLTestBool(pszValue);
}

/**********************************************************************
 *                       CSLFetchBoolean()
 **********************************************************************/

/** DEPRECATED.  Check for boolean key value.
 *
 * In a StringList of "Name=Value" pairs, look to see if there is a key
 * with the given name, and if it can be interpreted as being TRUE.  If
 * the key appears without any "=Value" portion it will be considered true.
 * If the value is NO, FALSE or 0 it will be considered FALSE otherwise
 * if the key appears in the list it will be considered TRUE.  If the key
 * doesn't appear at all, the indicated default value will be returned.
 *
 * @param papszStrList the string list to search.
 * @param pszKey the key value to look for (case insensitive).
 * @param bDefault the value to return if the key isn't found at all.
 *
 * @return TRUE or FALSE
 */

int CSLFetchBoolean(CSLConstList papszStrList, const char *pszKey, int bDefault)

{
    return CPLFetchBool(papszStrList, pszKey, CPL_TO_BOOL(bDefault));
}

/************************************************************************/
/*                     CSLFetchNameValueDefaulted()                     */
/************************************************************************/

/** Same as CSLFetchNameValue() but return pszDefault in case of no match */
const char *CSLFetchNameValueDef(CSLConstList papszStrList, const char *pszName,
                                 const char *pszDefault)

{
    const char *pszResult = CSLFetchNameValue(papszStrList, pszName);
    if (pszResult != nullptr)
        return pszResult;

    return pszDefault;
}

/**********************************************************************
 *                       CSLFetchNameValue()
 **********************************************************************/

/** In a StringList of "Name=Value" pairs, look for the
 * first value associated with the specified name.  The search is not
 * case sensitive.
 * ("Name:Value" pairs are also supported for backward compatibility
 * with older stuff.)
 *
 * Returns a reference to the value in the StringList that the caller
 * should not attempt to free.
 *
 * Returns NULL if the name is not found.
 */

const char *CSLFetchNameValue(CSLConstList papszStrList, const char *pszName)
{
    if (papszStrList == nullptr || pszName == nullptr)
        return nullptr;

    const size_t nLen = strlen(pszName);
    while (*papszStrList != nullptr)
    {
        if (EQUALN(*papszStrList, pszName, nLen) &&
            ((*papszStrList)[nLen] == '=' || (*papszStrList)[nLen] == ':'))
        {
            return (*papszStrList) + nLen + 1;
        }
        ++papszStrList;
    }
    return nullptr;
}

/************************************************************************/
/*                            CSLFindName()                             */
/************************************************************************/

/**
 * Find StringList entry with given key name.
 *
 * @param papszStrList the string list to search.
 * @param pszName the key value to look for (case insensitive).
 *
 * @return -1 on failure or the list index of the first occurrence
 * matching the given key.
 */

int CSLFindName(CSLConstList papszStrList, const char *pszName)
{
    if (papszStrList == nullptr || pszName == nullptr)
        return -1;

    const size_t nLen = strlen(pszName);
    int iIndex = 0;
    while (*papszStrList != nullptr)
    {
        if (EQUALN(*papszStrList, pszName, nLen) &&
            ((*papszStrList)[nLen] == '=' || (*papszStrList)[nLen] == ':'))
        {
            return iIndex;
        }
        ++iIndex;
        ++papszStrList;
    }
    return -1;
}

/************************************************************************/
/*                     CPLParseMemorySize()                             */
/************************************************************************/

/** Parse a memory size from a string.
 *
 * The string may indicate the units of the memory (e.g., "230k", "500 MB"),
 * using the prefixes "k", "m", or "g" in either lower or upper-case,
 * optionally followed by a "b" or "B". The string may alternatively specify
 * memory as a fraction of the usable RAM (e.g., "25%"). Spaces before the
 * number, between the number and the units, or after the units are ignored,
 * but other characters will cause a parsing failure. If the string cannot
 * be understood, the function will return CE_Failure.
 *
 * @param pszValue the string to parse
 * @param[out] pnValue the parsed size, converted to bytes (if unit was specified)
 * @param[out] pbUnitSpecified whether the string indicated the units
 *
 * @return CE_None on success, CE_Failure otherwise
 * @since 3.10
 */
CPLErr CPLParseMemorySize(const char *pszValue, GIntBig *pnValue,
                          bool *pbUnitSpecified)
{
    const char *start = pszValue;
    char *end = nullptr;

    // trim leading whitespace
    while (*start == ' ')
    {
        start++;
    }

    auto len = CPLStrnlen(start, 100);
    double value = CPLStrtodM(start, &end);
    const char *unit = nullptr;
    bool unitIsNotPercent = false;

    if (end == start)
    {
        CPLError(CE_Failure, CPLE_IllegalArg, "Received non-numeric value: %s",
                 pszValue);
        return CE_Failure;
    }

    if (value < 0 || !std::isfinite(value))
    {
        CPLError(CE_Failure, CPLE_IllegalArg,
                 "Memory size must be a positive number or zero.");
        return CE_Failure;
    }

    for (const char *c = end; c < start + len; c++)
    {
        if (unit == nullptr)
        {
            // check various suffixes and convert number into bytes
            if (*c == '%')
            {
                if (value < 0 || value > 100)
                {
                    CPLError(CE_Failure, CPLE_IllegalArg,
                             "Memory percentage must be between 0 and 100.");
                    return CE_Failure;
                }
                auto bytes = CPLGetUsablePhysicalRAM();
                if (bytes == 0)
                {
                    CPLError(CE_Failure, CPLE_NotSupported,
                             "Cannot determine usable physical RAM");
                    return CE_Failure;
                }
                value *= static_cast<double>(bytes / 100);
                unit = c;
            }
            else
            {
                switch (*c)
                {
                    case 'G':
                    case 'g':
                        value *= 1024;  // fall-through
                    case 'M':
                    case 'm':
                        value *= 1024;  // fall-through
                    case 'K':
                    case 'k':
                        value *= 1024;
                        unit = c;
                        unitIsNotPercent = true;
                        break;
                    case ' ':
                        break;
                    default:
                        CPLError(CE_Failure, CPLE_IllegalArg,
                                 "Unexpected value: %s", pszValue);
                        return CE_Failure;
                }
            }
        }
        else if (unitIsNotPercent && c == unit + 1 && (*c == 'b' || *c == 'B'))
        {
            // ignore 'B' or 'b' as part of unit
            continue;
        }
        else if (*c != ' ')
        {
            CPLError(CE_Failure, CPLE_IllegalArg, "Unexpected value: %s",
                     pszValue);
            return CE_Failure;
        }
    }

    *pnValue = static_cast<GIntBig>(value);
    if (pbUnitSpecified)
    {
        *pbUnitSpecified = (unit != nullptr);
    }
    return CE_None;
}

/**********************************************************************
 *                       CPLParseNameValue()
 **********************************************************************/

/**
 * Parse NAME=VALUE string into name and value components.
 *
 * Note that if ppszKey is non-NULL, the key (or name) portion will be
 * allocated using CPLMalloc(), and returned in that pointer.  It is the
 * applications responsibility to free this string, but the application should
 * not modify or free the returned value portion.
 *
 * This function also support "NAME:VALUE" strings and will strip white
 * space from around the delimiter when forming name and value strings.
 *
 * Eventually CSLFetchNameValue() and friends may be modified to use
 * CPLParseNameValue().
 *
 * @param pszNameValue string in "NAME=VALUE" format.
 * @param ppszKey optional pointer though which to return the name
 * portion.
 *
 * @return the value portion (pointing into original string).
 */

const char *CPLParseNameValue(const char *pszNameValue, char **ppszKey)
{
    for (int i = 0; pszNameValue[i] != '\0'; ++i)
    {
        if (pszNameValue[i] == '=' || pszNameValue[i] == ':')
        {
            const char *pszValue = pszNameValue + i + 1;
            while (*pszValue == ' ' || *pszValue == '\t')
                ++pszValue;

            if (ppszKey != nullptr)
            {
                *ppszKey = static_cast<char *>(CPLMalloc(i + 1));
                memcpy(*ppszKey, pszNameValue, i);
                (*ppszKey)[i] = '\0';
                while (i > 0 &&
                       ((*ppszKey)[i - 1] == ' ' || (*ppszKey)[i - 1] == '\t'))
                {
                    (*ppszKey)[i - 1] = '\0';
                    i--;
                }
            }

            return pszValue;
        }
    }

    return nullptr;
}

/**********************************************************************
 *                       CPLParseNameValueSep()
 **********************************************************************/
/**
 * Parse NAME<Sep>VALUE string into name and value components.
 *
 * This is derived directly from CPLParseNameValue() which will separate
 * on '=' OR ':', here chSep is required for specifying the separator
 * explicitly.
 *
 * @param pszNameValue string in "NAME=VALUE" format.
 * @param ppszKey optional pointer though which to return the name
 * portion.
 * @param chSep required single char separator
 * @return the value portion (pointing into original string).
 */

const char *CPLParseNameValueSep(const char *pszNameValue, char **ppszKey,
                                 char chSep)
{
    for (int i = 0; pszNameValue[i] != '\0'; ++i)
    {
        if (pszNameValue[i] == chSep)
        {
            const char *pszValue = pszNameValue + i + 1;
            while (*pszValue == ' ' || *pszValue == '\t')
                ++pszValue;

            if (ppszKey != nullptr)
            {
                *ppszKey = static_cast<char *>(CPLMalloc(i + 1));
                memcpy(*ppszKey, pszNameValue, i);
                (*ppszKey)[i] = '\0';
                while (i > 0 &&
                       ((*ppszKey)[i - 1] == ' ' || (*ppszKey)[i - 1] == '\t'))
                {
                    (*ppszKey)[i - 1] = '\0';
                    i--;
                }
            }

            return pszValue;
        }
    }

    return nullptr;
}

/**********************************************************************
 *                       CSLFetchNameValueMultiple()
 **********************************************************************/

/** In a StringList of "Name=Value" pairs, look for all the
 * values with the specified name.  The search is not case
 * sensitive.
 * ("Name:Value" pairs are also supported for backward compatibility
 * with older stuff.)
 *
 * Returns StringList with one entry for each occurrence of the
 * specified name.  The StringList should eventually be destroyed
 * by calling CSLDestroy().
 *
 * Returns NULL if the name is not found.
 */

char **CSLFetchNameValueMultiple(CSLConstList papszStrList, const char *pszName)
{
    if (papszStrList == nullptr || pszName == nullptr)
        return nullptr;

    const size_t nLen = strlen(pszName);
    char **papszValues = nullptr;
    while (*papszStrList != nullptr)
    {
        if (EQUALN(*papszStrList, pszName, nLen) &&
            ((*papszStrList)[nLen] == '=' || (*papszStrList)[nLen] == ':'))
        {
            papszValues = CSLAddString(papszValues, (*papszStrList) + nLen + 1);
        }
        ++papszStrList;
    }

    return papszValues;
}

/**********************************************************************
 *                       CSLAddNameValue()
 **********************************************************************/

/** Add a new entry to a StringList of "Name=Value" pairs,
 * ("Name:Value" pairs are also supported for backward compatibility
 * with older stuff.)
 *
 * This function does not check if a "Name=Value" pair already exists
 * for that name and can generate multiple entries for the same name.
 * Use CSLSetNameValue() if you want each name to have only one value.
 *
 * Returns the modified StringList.
 */

char **CSLAddNameValue(char **papszStrList, const char *pszName,
                       const char *pszValue)
{
    if (pszName == nullptr || pszValue == nullptr)
        return papszStrList;

    const size_t nLen = strlen(pszName) + strlen(pszValue) + 2;
    char *pszLine = static_cast<char *>(CPLMalloc(nLen));
    snprintf(pszLine, nLen, "%s=%s", pszName, pszValue);
    papszStrList = CSLAddString(papszStrList, pszLine);
    CPLFree(pszLine);

    return papszStrList;
}

/************************************************************************/
/*                          CSLSetNameValue()                           */
/************************************************************************/

/**
 * Assign value to name in StringList.
 *
 * Set the value for a given name in a StringList of "Name=Value" pairs
 * ("Name:Value" pairs are also supported for backward compatibility
 * with older stuff.)
 *
 * If there is already a value for that name in the list then the value
 * is changed, otherwise a new "Name=Value" pair is added.
 *
 * @param papszList the original list, the modified version is returned.
 * @param pszName the name to be assigned a value.  This should be a well
 * formed token (no spaces or very special characters).
 * @param pszValue the value to assign to the name.  This should not contain
 * any newlines (CR or LF) but is otherwise pretty much unconstrained.  If
 * NULL any corresponding value will be removed.
 *
 * @return modified StringList.
 */

char **CSLSetNameValue(char **papszList, const char *pszName,
                       const char *pszValue)
{
    if (pszName == nullptr)
        return papszList;

    size_t nLen = strlen(pszName);
    while (nLen > 0 && pszName[nLen - 1] == ' ')
        nLen--;
    char **papszPtr = papszList;
    while (papszPtr && *papszPtr != nullptr)
    {
        if (EQUALN(*papszPtr, pszName, nLen))
        {
            size_t i;
            for (i = nLen; (*papszPtr)[i] == ' '; ++i)
            {
            }
            if ((*papszPtr)[i] == '=' || (*papszPtr)[i] == ':')
            {
                // Found it.
                // Change the value... make sure to keep the ':' or '='.
                const char cSep = (*papszPtr)[i];

                CPLFree(*papszPtr);

                // If the value is NULL, remove this entry completely.
                if (pszValue == nullptr)
                {
                    while (papszPtr[1] != nullptr)
                    {
                        *papszPtr = papszPtr[1];
                        ++papszPtr;
                    }
                    *papszPtr = nullptr;
                }

                // Otherwise replace with new value.
                else
                {
                    const size_t nLen2 = strlen(pszName) + strlen(pszValue) + 2;
                    *papszPtr = static_cast<char *>(CPLMalloc(nLen2));
                    snprintf(*papszPtr, nLen2, "%s%c%s", pszName, cSep,
                             pszValue);
                }
                return papszList;
            }
        }
        ++papszPtr;
    }

    if (pszValue == nullptr)
        return papszList;

    // The name does not exist yet.  Create a new entry.
    return CSLAddNameValue(papszList, pszName, pszValue);
}

/************************************************************************/
/*                      CSLSetNameValueSeparator()                      */
/************************************************************************/

/**
 * Replace the default separator (":" or "=") with the passed separator
 * in the given name/value list.
 *
 * Note that if a separator other than ":" or "=" is used, the resulting
 * list will not be manipulable by the CSL name/value functions any more.
 *
 * The CPLParseNameValue() function is used to break the existing lines,
 * and it also strips white space from around the existing delimiter, thus
 * the old separator, and any white space will be replaced by the new
 * separator.  For formatting purposes it may be desirable to include some
 * white space in the new separator.  e.g. ": " or " = ".
 *
 * @param papszList the list to update.  Component strings may be freed
 * but the list array will remain at the same location.
 *
 * @param pszSeparator the new separator string to insert.
 */

void CSLSetNameValueSeparator(char **papszList, const char *pszSeparator)

{
    const int nLines = CSLCount(papszList);

    for (int iLine = 0; iLine < nLines; ++iLine)
    {
        char *pszKey = nullptr;
        const char *pszValue = CPLParseNameValue(papszList[iLine], &pszKey);
        if (pszValue == nullptr || pszKey == nullptr)
        {
            CPLFree(pszKey);
            continue;
        }

        char *pszNewLine = static_cast<char *>(CPLMalloc(
            strlen(pszValue) + strlen(pszKey) + strlen(pszSeparator) + 1));
        strcpy(pszNewLine, pszKey);
        strcat(pszNewLine, pszSeparator);
        strcat(pszNewLine, pszValue);
        CPLFree(papszList[iLine]);
        papszList[iLine] = pszNewLine;
        CPLFree(pszKey);
    }
}

/************************************************************************/
/*                          CPLEscapeString()                           */
/************************************************************************/

/**
 * Apply escaping to string to preserve special characters.
 *
 * This function will "escape" a variety of special characters
 * to make the string suitable to embed within a string constant
 * or to write within a text stream but in a form that can be
 * reconstituted to its original form.  The escaping will even preserve
 * zero bytes allowing preservation of raw binary data.
 *
 * CPLES_BackslashQuotable(0): This scheme turns a binary string into
 * a form suitable to be placed within double quotes as a string constant.
 * The backslash, quote, '\\0' and newline characters are all escaped in
 * the usual C style.
 *
 * CPLES_XML(1): This scheme converts the '<', '>', '"' and '&' characters into
 * their XML/HTML equivalent (&lt;, &gt;, &quot; and &amp;) making a string safe
 * to embed as CDATA within an XML element.  The '\\0' is not escaped and
 * should not be included in the input.
 *
 * CPLES_URL(2): Everything except alphanumerics and the characters
 * '$', '-', '_', '.', '+', '!', '*', ''', '(', ')' and ',' (see RFC1738) are
 * converted to a percent followed by a two digit hex encoding of the character
 * (leading zero supplied if needed).  This is the mechanism used for encoding
 * values to be passed in URLs.
 *
 * CPLES_SQL(3): All single quotes are replaced with two single quotes.
 * Suitable for use when constructing literal values for SQL commands where
 * the literal will be enclosed in single quotes.
 *
 * CPLES_CSV(4): If the values contains commas, semicolons, tabs, double quotes,
 * or newlines it placed in double quotes, and double quotes in the value are
 * doubled. Suitable for use when constructing field values for .csv files.
 * Note that CPLUnescapeString() currently does not support this format, only
 * CPLEscapeString().  See cpl_csv.cpp for CSV parsing support.
 *
 * CPLES_SQLI(7): All double quotes are replaced with two double quotes.
 * Suitable for use when constructing identifiers for SQL commands where
 * the literal will be enclosed in double quotes.
 *
 * @param pszInput the string to escape.
 * @param nLength The number of bytes of data to preserve.  If this is -1
 * the strlen(pszString) function will be used to compute the length.
 * @param nScheme the encoding scheme to use.
 *
 * @return an escaped, zero terminated string that should be freed with
 * CPLFree() when no longer needed.
 */

char *CPLEscapeString(const char *pszInput, int nLength, int nScheme)
{
    const size_t szLength =
        (nLength < 0) ? strlen(pszInput) : static_cast<size_t>(nLength);
#define nLength no_longer_use_me

    size_t nSizeAlloc = 1;
#if SIZEOF_VOIDP < 8
    bool bWrapAround = false;
    const auto IncSizeAlloc = [&nSizeAlloc, &bWrapAround](size_t inc)
    {
        constexpr size_t SZ_MAX = std::numeric_limits<size_t>::max();
        if (nSizeAlloc > SZ_MAX - inc)
        {
            bWrapAround = true;
            nSizeAlloc = 0;
        }
        nSizeAlloc += inc;
    };
#else
    const auto IncSizeAlloc = [&nSizeAlloc](size_t inc) { nSizeAlloc += inc; };
#endif

    if (nScheme == CPLES_BackslashQuotable)
    {
        for (size_t iIn = 0; iIn < szLength; iIn++)
        {
            if (pszInput[iIn] == '\0' || pszInput[iIn] == '\n' ||
                pszInput[iIn] == '"' || pszInput[iIn] == '\\')
                IncSizeAlloc(2);
            else
                IncSizeAlloc(1);
        }
    }
    else if (nScheme == CPLES_XML || nScheme == CPLES_XML_BUT_QUOTES)
    {
        for (size_t iIn = 0; iIn < szLength; ++iIn)
        {
            if (pszInput[iIn] == '<')
            {
                IncSizeAlloc(4);
            }
            else if (pszInput[iIn] == '>')
            {
                IncSizeAlloc(4);
            }
            else if (pszInput[iIn] == '&')
            {
                IncSizeAlloc(5);
            }
            else if (pszInput[iIn] == '"' && nScheme != CPLES_XML_BUT_QUOTES)
            {
                IncSizeAlloc(6);
            }
            // Python 2 does not display the UTF-8 character corresponding
            // to the byte-order mark (BOM), so escape it.
            else if ((reinterpret_cast<const unsigned char *>(pszInput))[iIn] ==
                         0xEF &&
                     (reinterpret_cast<const unsigned char *>(
                         pszInput))[iIn + 1] == 0xBB &&
                     (reinterpret_cast<const unsigned char *>(
                         pszInput))[iIn + 2] == 0xBF)
            {
                IncSizeAlloc(8);
                iIn += 2;
            }
            else if ((reinterpret_cast<const unsigned char *>(pszInput))[iIn] <
                         0x20 &&
                     pszInput[iIn] != 0x9 && pszInput[iIn] != 0xA &&
                     pszInput[iIn] != 0xD)
            {
                // These control characters are unrepresentable in XML format,
                // so we just drop them.  #4117
            }
            else
            {
                IncSizeAlloc(1);
            }
        }
    }
    else if (nScheme == CPLES_URL)  // Untested at implementation.
    {
        for (size_t iIn = 0; iIn < szLength; ++iIn)
        {
            if ((pszInput[iIn] >= 'a' && pszInput[iIn] <= 'z') ||
                (pszInput[iIn] >= 'A' && pszInput[iIn] <= 'Z') ||
                (pszInput[iIn] >= '0' && pszInput[iIn] <= '9') ||
                pszInput[iIn] == '$' || pszInput[iIn] == '-' ||
                pszInput[iIn] == '_' || pszInput[iIn] == '.' ||
                pszInput[iIn] == '+' || pszInput[iIn] == '!' ||
                pszInput[iIn] == '*' || pszInput[iIn] == '\'' ||
                pszInput[iIn] == '(' || pszInput[iIn] == ')' ||
                pszInput[iIn] == ',')
            {
                IncSizeAlloc(1);
            }
            else
            {
                IncSizeAlloc(3);
            }
        }
    }
    else if (nScheme == CPLES_SQL || nScheme == CPLES_SQLI)
    {
        const char chQuote = nScheme == CPLES_SQL ? '\'' : '\"';
        for (size_t iIn = 0; iIn < szLength; ++iIn)
        {
            if (pszInput[iIn] == chQuote)
            {
                IncSizeAlloc(2);
            }
            else
            {
                IncSizeAlloc(1);
            }
        }
    }
    else if (nScheme == CPLES_CSV || nScheme == CPLES_CSV_FORCE_QUOTING)
    {
        if (nScheme == CPLES_CSV && strcspn(pszInput, "\",;\t\n\r") == szLength)
        {
            char *pszOutput =
                static_cast<char *>(VSI_MALLOC_VERBOSE(szLength + 1));
            if (pszOutput == nullptr)
                return nullptr;
            memcpy(pszOutput, pszInput, szLength + 1);
            return pszOutput;
        }
        else
        {
            IncSizeAlloc(1);
            for (size_t iIn = 0; iIn < szLength; ++iIn)
            {
                if (pszInput[iIn] == '\"')
                {
                    IncSizeAlloc(2);
                }
                else
                    IncSizeAlloc(1);
            }
            IncSizeAlloc(1);
        }
    }
    else
    {
        CPLError(CE_Failure, CPLE_AppDefined,
                 "Undefined escaping scheme (%d) in CPLEscapeString()",
                 nScheme);
        return CPLStrdup("");
    }

#if SIZEOF_VOIDP < 8
    if (bWrapAround)
    {
        CPLError(CE_Failure, CPLE_OutOfMemory,
                 "Out of memory in CPLEscapeString()");
        return nullptr;
    }
#endif

    char *pszOutput = static_cast<char *>(VSI_MALLOC_VERBOSE(nSizeAlloc));
    if (pszOutput == nullptr)
        return nullptr;

    size_t iOut = 0;

    if (nScheme == CPLES_BackslashQuotable)
    {
        for (size_t iIn = 0; iIn < szLength; iIn++)
        {
            if (pszInput[iIn] == '\0')
            {
                pszOutput[iOut++] = '\\';
                pszOutput[iOut++] = '0';
            }
            else if (pszInput[iIn] == '\n')
            {
                pszOutput[iOut++] = '\\';
                pszOutput[iOut++] = 'n';
            }
            else if (pszInput[iIn] == '"')
            {
                pszOutput[iOut++] = '\\';
                pszOutput[iOut++] = '\"';
            }
            else if (pszInput[iIn] == '\\')
            {
                pszOutput[iOut++] = '\\';
                pszOutput[iOut++] = '\\';
            }
            else
                pszOutput[iOut++] = pszInput[iIn];
        }
        pszOutput[iOut++] = '\0';
    }
    else if (nScheme == CPLES_XML || nScheme == CPLES_XML_BUT_QUOTES)
    {
        for (size_t iIn = 0; iIn < szLength; ++iIn)
        {
            if (pszInput[iIn] == '<')
            {
                pszOutput[iOut++] = '&';
                pszOutput[iOut++] = 'l';
                pszOutput[iOut++] = 't';
                pszOutput[iOut++] = ';';
            }
            else if (pszInput[iIn] == '>')
            {
                pszOutput[iOut++] = '&';
                pszOutput[iOut++] = 'g';
                pszOutput[iOut++] = 't';
                pszOutput[iOut++] = ';';
            }
            else if (pszInput[iIn] == '&')
            {
                pszOutput[iOut++] = '&';
                pszOutput[iOut++] = 'a';
                pszOutput[iOut++] = 'm';
                pszOutput[iOut++] = 'p';
                pszOutput[iOut++] = ';';
            }
            else if (pszInput[iIn] == '"' && nScheme != CPLES_XML_BUT_QUOTES)
            {
                pszOutput[iOut++] = '&';
                pszOutput[iOut++] = 'q';
                pszOutput[iOut++] = 'u';
                pszOutput[iOut++] = 'o';
                pszOutput[iOut++] = 't';
                pszOutput[iOut++] = ';';
            }
            // Python 2 does not display the UTF-8 character corresponding
            // to the byte-order mark (BOM), so escape it.
            else if ((reinterpret_cast<const unsigned char *>(pszInput))[iIn] ==
                         0xEF &&
                     (reinterpret_cast<const unsigned char *>(
                         pszInput))[iIn + 1] == 0xBB &&
                     (reinterpret_cast<const unsigned char *>(
                         pszInput))[iIn + 2] == 0xBF)
            {
                pszOutput[iOut++] = '&';
                pszOutput[iOut++] = '#';
                pszOutput[iOut++] = 'x';
                pszOutput[iOut++] = 'F';
                pszOutput[iOut++] = 'E';
                pszOutput[iOut++] = 'F';
                pszOutput[iOut++] = 'F';
                pszOutput[iOut++] = ';';
                iIn += 2;
            }
            else if ((reinterpret_cast<const unsigned char *>(pszInput))[iIn] <
                         0x20 &&
                     pszInput[iIn] != 0x9 && pszInput[iIn] != 0xA &&
                     pszInput[iIn] != 0xD)
            {
                // These control characters are unrepresentable in XML format,
                // so we just drop them.  #4117
            }
            else
            {
                pszOutput[iOut++] = pszInput[iIn];
            }
        }
        pszOutput[iOut++] = '\0';
    }
    else if (nScheme == CPLES_URL)  // Untested at implementation.
    {
        for (size_t iIn = 0; iIn < szLength; ++iIn)
        {
            if ((pszInput[iIn] >= 'a' && pszInput[iIn] <= 'z') ||
                (pszInput[iIn] >= 'A' && pszInput[iIn] <= 'Z') ||
                (pszInput[iIn] >= '0' && pszInput[iIn] <= '9') ||
                pszInput[iIn] == '$' || pszInput[iIn] == '-' ||
                pszInput[iIn] == '_' || pszInput[iIn] == '.' ||
                pszInput[iIn] == '+' || pszInput[iIn] == '!' ||
                pszInput[iIn] == '*' || pszInput[iIn] == '\'' ||
                pszInput[iIn] == '(' || pszInput[iIn] == ')' ||
                pszInput[iIn] == ',')
            {
                pszOutput[iOut++] = pszInput[iIn];
            }
            else
            {
                snprintf(pszOutput + iOut, nSizeAlloc - iOut, "%%%02X",
                         static_cast<unsigned char>(pszInput[iIn]));
                iOut += 3;
            }
        }
        pszOutput[iOut++] = '\0';
    }
    else if (nScheme == CPLES_SQL || nScheme == CPLES_SQLI)
    {
        const char chQuote = nScheme == CPLES_SQL ? '\'' : '\"';
        for (size_t iIn = 0; iIn < szLength; ++iIn)
        {
            if (pszInput[iIn] == chQuote)
            {
                pszOutput[iOut++] = chQuote;
                pszOutput[iOut++] = chQuote;
            }
            else
            {
                pszOutput[iOut++] = pszInput[iIn];
            }
        }
        pszOutput[iOut++] = '\0';
    }
    else if (nScheme == CPLES_CSV || nScheme == CPLES_CSV_FORCE_QUOTING)
    {
        pszOutput[iOut++] = '\"';

        for (size_t iIn = 0; iIn < szLength; ++iIn)
        {
            if (pszInput[iIn] == '\"')
            {
                pszOutput[iOut++] = '\"';
                pszOutput[iOut++] = '\"';
            }
            else
                pszOutput[iOut++] = pszInput[iIn];
        }
        pszOutput[iOut++] = '\"';
        pszOutput[iOut++] = '\0';
    }

    return pszOutput;
#undef nLength
}

/************************************************************************/
/*                         CPLUnescapeString()                          */
/************************************************************************/

/**
 * Unescape a string.
 *
 * This function does the opposite of CPLEscapeString().  Given a string
 * with special values escaped according to some scheme, it will return a
 * new copy of the string returned to its original form.
 *
 * @param pszInput the input string.  This is a zero terminated string.
 * @param pnLength location to return the length of the unescaped string,
 * which may in some cases include embedded '\\0' characters.
 * @param nScheme the escaped scheme to undo (see CPLEscapeString() for a
 * list).  Does not yet support CSV.
 *
 * @return a copy of the unescaped string that should be freed by the
 * application using CPLFree() when no longer needed.
 */

CPL_NOSANITIZE_UNSIGNED_INT_OVERFLOW
char *CPLUnescapeString(const char *pszInput, int *pnLength, int nScheme)

{
    int iOut = 0;

    // TODO: Why times 4?
    char *pszOutput = static_cast<char *>(CPLMalloc(4 * strlen(pszInput) + 1));
    pszOutput[0] = '\0';

    if (nScheme == CPLES_BackslashQuotable)
    {
        for (int iIn = 0; pszInput[iIn] != '\0'; ++iIn)
        {
            if (pszInput[iIn] == '\\')
            {
                ++iIn;
                if (pszInput[iIn] == '\0')
                    break;
                if (pszInput[iIn] == 'n')
                    pszOutput[iOut++] = '\n';
                else if (pszInput[iIn] == '0')
                    pszOutput[iOut++] = '\0';
                else
                    pszOutput[iOut++] = pszInput[iIn];
            }
            else
            {
                pszOutput[iOut++] = pszInput[iIn];
            }
        }
    }
    else if (nScheme == CPLES_XML || nScheme == CPLES_XML_BUT_QUOTES)
    {
        char ch = '\0';
        for (int iIn = 0; (ch = pszInput[iIn]) != '\0'; ++iIn)
        {
            if (ch != '&')
            {
                pszOutput[iOut++] = ch;
            }
            else if (STARTS_WITH_CI(pszInput + iIn, "&lt;"))
            {
                pszOutput[iOut++] = '<';
                iIn += 3;
            }
            else if (STARTS_WITH_CI(pszInput + iIn, "&gt;"))
            {
                pszOutput[iOut++] = '>';
                iIn += 3;
            }
            else if (STARTS_WITH_CI(pszInput + iIn, "&amp;"))
            {
                pszOutput[iOut++] = '&';
                iIn += 4;
            }
            else if (STARTS_WITH_CI(pszInput + iIn, "&apos;"))
            {
                pszOutput[iOut++] = '\'';
                iIn += 5;
            }
            else if (STARTS_WITH_CI(pszInput + iIn, "&quot;"))
            {
                pszOutput[iOut++] = '"';
                iIn += 5;
            }
            else if (STARTS_WITH_CI(pszInput + iIn, "&#x"))
            {
                wchar_t anVal[2] = {0, 0};
                iIn += 3;

                unsigned int nVal = 0;
                while (true)
                {
                    ch = pszInput[iIn++];
                    if (ch >= 'a' && ch <= 'f')
                        nVal = nVal * 16U +
                               static_cast<unsigned int>(ch - 'a' + 10);
                    else if (ch >= 'A' && ch <= 'F')
                        nVal = nVal * 16U +
                               static_cast<unsigned int>(ch - 'A' + 10);
                    else if (ch >= '0' && ch <= '9')
                        nVal = nVal * 16U + static_cast<unsigned int>(ch - '0');
                    else
                        break;
                }
                anVal[0] = static_cast<wchar_t>(nVal);
                if (ch != ';')
                    break;
                iIn--;

                char *pszUTF8 =
                    CPLRecodeFromWChar(anVal, "WCHAR_T", CPL_ENC_UTF8);
                int nLen = static_cast<int>(strlen(pszUTF8));
                memcpy(pszOutput + iOut, pszUTF8, nLen);
                CPLFree(pszUTF8);
                iOut += nLen;
            }
            else if (STARTS_WITH_CI(pszInput + iIn, "&#"))
            {
                wchar_t anVal[2] = {0, 0};
                iIn += 2;

                unsigned int nVal = 0;
                while (true)
                {
                    ch = pszInput[iIn++];
                    if (ch >= '0' && ch <= '9')
                        nVal = nVal * 10U + static_cast<unsigned int>(ch - '0');
                    else
                        break;
                }
                anVal[0] = static_cast<wchar_t>(nVal);
                if (ch != ';')
                    break;
                iIn--;

                char *pszUTF8 =
                    CPLRecodeFromWChar(anVal, "WCHAR_T", CPL_ENC_UTF8);
                const int nLen = static_cast<int>(strlen(pszUTF8));
                memcpy(pszOutput + iOut, pszUTF8, nLen);
                CPLFree(pszUTF8);
                iOut += nLen;
            }
            else
            {
                // Illegal escape sequence.
                CPLDebug("CPL",
                         "Error unescaping CPLES_XML text, '&' character "
                         "followed by unhandled escape sequence.");
                break;
            }
        }
    }
    else if (nScheme == CPLES_URL)
    {
        for (int iIn = 0; pszInput[iIn] != '\0'; ++iIn)
        {
            if (pszInput[iIn] == '%' && pszInput[iIn + 1] != '\0' &&
                pszInput[iIn + 2] != '\0')
            {
                int nHexChar = 0;

                if (pszInput[iIn + 1] >= 'A' && pszInput[iIn + 1] <= 'F')
                    nHexChar += 16 * (pszInput[iIn + 1] - 'A' + 10);
                else if (pszInput[iIn + 1] >= 'a' && pszInput[iIn + 1] <= 'f')
                    nHexChar += 16 * (pszInput[iIn + 1] - 'a' + 10);
                else if (pszInput[iIn + 1] >= '0' && pszInput[iIn + 1] <= '9')
                    nHexChar += 16 * (pszInput[iIn + 1] - '0');
                else
                    CPLDebug("CPL",
                             "Error unescaping CPLES_URL text, percent not "
                             "followed by two hex digits.");

                if (pszInput[iIn + 2] >= 'A' && pszInput[iIn + 2] <= 'F')
                    nHexChar += pszInput[iIn + 2] - 'A' + 10;
                else if (pszInput[iIn + 2] >= 'a' && pszInput[iIn + 2] <= 'f')
                    nHexChar += pszInput[iIn + 2] - 'a' + 10;
                else if (pszInput[iIn + 2] >= '0' && pszInput[iIn + 2] <= '9')
                    nHexChar += pszInput[iIn + 2] - '0';
                else
                    CPLDebug("CPL",
                             "Error unescaping CPLES_URL text, percent not "
                             "followed by two hex digits.");

                pszOutput[iOut++] = static_cast<char>(nHexChar);
                iIn += 2;
            }
            else if (pszInput[iIn] == '+')
            {
                pszOutput[iOut++] = ' ';
            }
            else
            {
                pszOutput[iOut++] = pszInput[iIn];
            }
        }
    }
    else if (nScheme == CPLES_SQL || nScheme == CPLES_SQLI)
    {
        char szQuote = nScheme == CPLES_SQL ? '\'' : '\"';
        for (int iIn = 0; pszInput[iIn] != '\0'; ++iIn)
        {
            if (pszInput[iIn] == szQuote && pszInput[iIn + 1] == szQuote)
            {
                ++iIn;
                pszOutput[iOut++] = pszInput[iIn];
            }
            else
            {
                pszOutput[iOut++] = pszInput[iIn];
            }
        }
    }
    else if (nScheme == CPLES_CSV)
    {
        CPLError(CE_Fatal, CPLE_NotSupported,
                 "CSV Unescaping not yet implemented.");
    }
    else
    {
        CPLError(CE_Fatal, CPLE_NotSupported, "Unknown escaping style.");
    }

    pszOutput[iOut] = '\0';

    if (pnLength != nullptr)
        *pnLength = iOut;

    return pszOutput;
}

/************************************************************************/
/*                           CPLBinaryToHex()                           */
/************************************************************************/

/**
 * Binary to hexadecimal translation.
 *
 * @param nBytes number of bytes of binary data in pabyData.
 * @param pabyData array of data bytes to translate.
 *
 * @return hexadecimal translation, zero terminated.  Free with CPLFree().
 */

char *CPLBinaryToHex(int nBytes, const GByte *pabyData)

{
    CPLAssert(nBytes >= 0);
    char *pszHex = static_cast<char *>(
        VSI_MALLOC_VERBOSE(static_cast<size_t>(nBytes) * 2 + 1));
    if (!pszHex)
    {
        pszHex = CPLStrdup("");
        return pszHex;
    }
    pszHex[nBytes * 2] = '\0';

    constexpr char achHex[] = "0123456789ABCDEF";

    for (size_t i = 0; i < static_cast<size_t>(nBytes); ++i)
    {
        const int nLow = pabyData[i] & 0x0f;
        const int nHigh = (pabyData[i] & 0xf0) >> 4;

        pszHex[i * 2] = achHex[nHigh];
        pszHex[i * 2 + 1] = achHex[nLow];
    }

    return pszHex;
}

/************************************************************************/
/*                           CPLHexToBinary()                           */
/************************************************************************/

constexpr unsigned char hex2char[256] = {
    // Not Hex characters.
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    // 0-9
    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 0, 0, 0, 0, 0, 0,
    // A-F
    0, 10, 11, 12, 13, 14, 15, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    // Not Hex characters.
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    // a-f
    0, 10, 11, 12, 13, 14, 15, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0,
    // Not Hex characters (upper 128 characters).
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0};

/**
 * Hexadecimal to binary translation
 *
 * @param pszHex the input hex encoded string.
 * @param pnBytes the returned count of decoded bytes placed here.
 *
 * @return returns binary buffer of data - free with CPLFree().
 */

GByte *CPLHexToBinary(const char *pszHex, int *pnBytes)
{
    const GByte *pabyHex = reinterpret_cast<const GByte *>(pszHex);
    const size_t nHexLen = strlen(pszHex);

    GByte *pabyWKB = static_cast<GByte *>(CPLMalloc(nHexLen / 2 + 2));

    for (size_t i = 0; i < nHexLen / 2; ++i)
    {
        const unsigned char h1 = hex2char[pabyHex[2 * i]];
        const unsigned char h2 = hex2char[pabyHex[2 * i + 1]];

        // First character is high bits, second is low bits.
        pabyWKB[i] = static_cast<GByte>((h1 << 4) | h2);
    }
    pabyWKB[nHexLen / 2] = 0;
    *pnBytes = static_cast<int>(nHexLen / 2);

    return pabyWKB;
}

/************************************************************************/
/*                         CPLGetValueType()                            */
/************************************************************************/

/**
 * Detect the type of the value contained in a string, whether it is
 * a real, an integer or a string
 * Leading and trailing spaces are skipped in the analysis.
 *
 * Note: in the context of this function, integer must be understood in a
 * broad sense. It does not mean that the value can fit into a 32 bit integer
 * for example. It might be larger.
 *
 * @param pszValue the string to analyze
 *
 * @return returns the type of the value contained in the string.
 */

CPLValueType CPLGetValueType(const char *pszValue)
{
    // Doubles : "+25.e+3", "-25.e-3", "25.e3", "25e3", " 25e3 "
    // Not doubles: "25e 3", "25e.3", "-2-5e3", "2-5e3", "25.25.3", "-3d", "d1"
    //              "XXeYYYYYYYYYYYYYYYYYYY" that evaluates to infinity

    if (pszValue == nullptr)
        return CPL_VALUE_STRING;

    const char *pszValueInit = pszValue;

    // Skip leading spaces.
    while (isspace(static_cast<unsigned char>(*pszValue)))
        ++pszValue;

    if (*pszValue == '\0')
        return CPL_VALUE_STRING;

    // Skip leading + or -.
    if (*pszValue == '+' || *pszValue == '-')
        ++pszValue;

    constexpr char DIGIT_ZERO = '0';
    if (pszValue[0] == DIGIT_ZERO && pszValue[1] != '\0' && pszValue[1] != '.')
        return CPL_VALUE_STRING;

    bool bFoundDot = false;
    bool bFoundExponent = false;
    bool bIsLastCharExponent = false;
    bool bIsReal = false;
    const char *pszAfterExponent = nullptr;
    bool bFoundMantissa = false;

    for (; *pszValue != '\0'; ++pszValue)
    {
        if (isdigit(static_cast<unsigned char>(*pszValue)))
        {
            bIsLastCharExponent = false;
            bFoundMantissa = true;
        }
        else if (isspace(static_cast<unsigned char>(*pszValue)))
        {
            const char *pszTmp = pszValue;
            while (isspace(static_cast<unsigned char>(*pszTmp)))
                ++pszTmp;
            if (*pszTmp == 0)
                break;
            else
                return CPL_VALUE_STRING;
        }
        else if (*pszValue == '-' || *pszValue == '+')
        {
            if (bIsLastCharExponent)
            {
                // Do nothing.
            }
            else
            {
                return CPL_VALUE_STRING;
            }
            bIsLastCharExponent = false;
        }
        else if (*pszValue == '.')
        {
            bIsReal = true;
            if (!bFoundDot && !bIsLastCharExponent)
                bFoundDot = true;
            else
                return CPL_VALUE_STRING;
            bIsLastCharExponent = false;
        }
        else if (*pszValue == 'D' || *pszValue == 'd' || *pszValue == 'E' ||
                 *pszValue == 'e')
        {
            if (!bFoundMantissa)
                return CPL_VALUE_STRING;
            if (!(pszValue[1] == '+' || pszValue[1] == '-' ||
                  isdigit(static_cast<unsigned char>(pszValue[1]))))
                return CPL_VALUE_STRING;

            bIsReal = true;
            if (!bFoundExponent)
                bFoundExponent = true;
            else
                return CPL_VALUE_STRING;
            pszAfterExponent = pszValue + 1;
            bIsLastCharExponent = true;
        }
        else
        {
            return CPL_VALUE_STRING;
        }
    }

    if (bIsReal && pszAfterExponent && strlen(pszAfterExponent) > 3)
    {
        // cppcheck-suppress unreadVariable
        const double dfVal = CPLAtof(pszValueInit);
        if (std::isinf(dfVal))
            return CPL_VALUE_STRING;
    }

    return bIsReal ? CPL_VALUE_REAL : CPL_VALUE_INTEGER;
}

/************************************************************************/
/*                              CPLStrlcpy()                            */
/************************************************************************/

/**
 * Copy source string to a destination buffer.
 *
 * This function ensures that the destination buffer is always NUL terminated
 * (provided that its length is at least 1).
 *
 * This function is designed to be a safer, more consistent, and less error
 * prone replacement for strncpy. Its contract is identical to libbsd's strlcpy.
 *
 * Truncation can be detected by testing if the return value of CPLStrlcpy
 * is greater or equal to nDestSize.

\verbatim
char szDest[5] = {};
if( CPLStrlcpy(szDest, "abcde", sizeof(szDest)) >= sizeof(szDest) )
    fprintf(stderr, "truncation occurred !\n");
\endverbatim

 * @param pszDest   destination buffer
 * @param pszSrc    source string. Must be NUL terminated
 * @param nDestSize size of destination buffer (including space for the NUL
 *     terminator character)
 *
 * @return the length of the source string (=strlen(pszSrc))
 *
 * @since GDAL 1.7.0
 */
size_t CPLStrlcpy(char *pszDest, const char *pszSrc, size_t nDestSize)
{
    if (nDestSize == 0)
        return strlen(pszSrc);

    char *pszDestIter = pszDest;
    const char *pszSrcIter = pszSrc;

    --nDestSize;
    while (nDestSize != 0 && *pszSrcIter != '\0')
    {
        *pszDestIter = *pszSrcIter;
        ++pszDestIter;
        ++pszSrcIter;
        --nDestSize;
    }
    *pszDestIter = '\0';
    return pszSrcIter - pszSrc + strlen(pszSrcIter);
}

/************************************************************************/
/*                              CPLStrlcat()                            */
/************************************************************************/

/**
 * Appends a source string to a destination buffer.
 *
 * This function ensures that the destination buffer is always NUL terminated
 * (provided that its length is at least 1 and that there is at least one byte
 * free in pszDest, that is to say strlen(pszDest_before) < nDestSize)
 *
 * This function is designed to be a safer, more consistent, and less error
 * prone replacement for strncat. Its contract is identical to libbsd's strlcat.
 *
 * Truncation can be detected by testing if the return value of CPLStrlcat
 * is greater or equal to nDestSize.

\verbatim
char szDest[5] = {};
CPLStrlcpy(szDest, "ab", sizeof(szDest));
if( CPLStrlcat(szDest, "cde", sizeof(szDest)) >= sizeof(szDest) )
    fprintf(stderr, "truncation occurred !\n");
\endverbatim

 * @param pszDest   destination buffer. Must be NUL terminated before
 *         running CPLStrlcat
 * @param pszSrc    source string. Must be NUL terminated
 * @param nDestSize size of destination buffer (including space for the
 *         NUL terminator character)
 *
 * @return the theoretical length of the destination string after concatenation
 *         (=strlen(pszDest_before) + strlen(pszSrc)).
 *         If strlen(pszDest_before) >= nDestSize, then it returns
 *         nDestSize + strlen(pszSrc)
 *
 * @since GDAL 1.7.0
 */
size_t CPLStrlcat(char *pszDest, const char *pszSrc, size_t nDestSize)
{
    char *pszDestIter = pszDest;

    while (nDestSize != 0 && *pszDestIter != '\0')
    {
        ++pszDestIter;
        --nDestSize;
    }

    return pszDestIter - pszDest + CPLStrlcpy(pszDestIter, pszSrc, nDestSize);
}

/************************************************************************/
/*                              CPLStrnlen()                            */
/************************************************************************/

/**
 * Returns the length of a NUL terminated string by reading at most
 * the specified number of bytes.
 *
 * The CPLStrnlen() function returns min(strlen(pszStr), nMaxLen).
 * Only the first nMaxLen bytes of the string will be read. Useful to
 * test if a string contains at least nMaxLen characters without reading
 * the full string up to the NUL terminating character.
 *
 * @param pszStr    a NUL terminated string
 * @param nMaxLen   maximum number of bytes to read in pszStr
 *
 * @return strlen(pszStr) if the length is lesser than nMaxLen, otherwise
 * nMaxLen if the NUL character has not been found in the first nMaxLen bytes.
 *
 * @since GDAL 1.7.0
 */

size_t CPLStrnlen(const char *pszStr, size_t nMaxLen)
{
    size_t nLen = 0;
    while (nLen < nMaxLen && *pszStr != '\0')
    {
        ++nLen;
        ++pszStr;
    }
    return nLen;
}

/************************************************************************/
/*                            CSLParseCommandLine()                     */
/************************************************************************/

/**
 * Tokenize command line arguments in a list of strings.
 *
 * @param pszCommandLine  command line
 *
 * @return NULL terminated list of strings to free with CSLDestroy()
 *
 * @since GDAL 2.1
 */
char **CSLParseCommandLine(const char *pszCommandLine)
{
    return CSLTokenizeString(pszCommandLine);
}

/************************************************************************/
/*                              CPLToupper()                            */
/************************************************************************/

/** Converts a (ASCII) lowercase character to uppercase.
 *
 * Same as standard toupper(), except that it is not locale sensitive.
 *
 * @since GDAL 3.9
 */
int CPLToupper(int c)
{
    return (c >= 'a' && c <= 'z') ? (c - 'a' + 'A') : c;
}

/************************************************************************/
/*                              CPLTolower()                            */
/************************************************************************/

/** Converts a (ASCII) uppercase character to lowercase.
 *
 * Same as standard tolower(), except that it is not locale sensitive.
 *
 * @since GDAL 3.9
 */
int CPLTolower(int c)
{
    return (c >= 'A' && c <= 'Z') ? (c - 'A' + 'a') : c;
}

/************************************************************************/
/*                      CPLRemoveSQLComments()                          */
/************************************************************************/

/** Remove SQL comments from a string
 *
 * @param osInput Input string.
 * @since GDAL 3.11
 */
std::string CPLRemoveSQLComments(const std::string &osInput)
{
    const CPLStringList aosLines(
        CSLTokenizeStringComplex(osInput.c_str(), "\r\n", FALSE, FALSE));
    std::string osSQL;
    for (const char *pszLine : aosLines)
    {
        char chQuote = 0;
        int i = 0;
        for (; pszLine[i] != '\0'; ++i)
        {
            if (chQuote)
            {
                if (pszLine[i] == chQuote)
                {
                    // Deal with escaped quote character which is repeated,
                    // so 'foo''bar' or "foo""bar"
                    if (pszLine[i + 1] == chQuote)
                    {
                        i++;
                    }
                    else
                    {
                        chQuote = 0;
                    }
                }
            }
            else if (pszLine[i] == '\'' || pszLine[i] == '"')
            {
                chQuote = pszLine[i];
            }
            else if (pszLine[i] == '-' && pszLine[i + 1] == '-')
            {
                break;
            }
        }
        if (i > 0)
        {
            if (!osSQL.empty())
                osSQL += ' ';
            osSQL.append(pszLine, i);
        }
    }
    return osSQL;
}