Free, open-source file converter. Convert 263 formats offline by drag & drop — images, PDF, Word, Excel, PowerPoint, HWP/HWPX, video, audio, e-books and archives. No upload, no Office needed.
C# audio-converteravaloniabatch-conversionconvertercsharpdocument-converterdocx-to-pdfdotnetdrag-and-dropfile-conversionfile-converterheic-converterhwphwpximage-converterlinuxmacospdf-convertervideo-converterwindows
// Repairs the one mistake that makes many hand-written and AI-generated SVGs unreadable: an "&" that isn't an entity
// ("A & B" in a <text>, a URL query in an href) and HTML entities XML doesn't know ( , ©). Browsers forgive
// some of it; the XML parser Svg.Skia uses stops at the first one (issue #120).
using System.Net;
using System.Text;
using System.Text.RegularExpressions;
namespace Filee.Engines.Vector;
internal static partial class SvgXmlRepair
{
/// <summary>The entities every XML document has; anything else must be declared in the DTD.</summary>
private static readonly HashSet<string> XmlEntities = ["amp", "lt", "gt", "quot", "apos"];
/// <summary>
/// Returns <paramref name="svg"/> with every "&" that doesn't start a character reference or a known entity
/// escaped, and HTML named entities turned into character references. Entities declared in the document's DTD
/// (Illustrator's &ns_svg;), the DOCTYPE, comments and CDATA sections are left alone. Null when nothing needed repair, or
/// when the bytes aren't UTF-8 (another declared encoding is read as it is).
/// </summary>
public static byte[]? FixAmpersands(byte[] svg)
{
if (svg.AsSpan().IndexOf((byte)'&') < 0)
return null;
string text;
try
{
text = new UTF8Encoding(false, throwOnInvalidBytes: true).GetString(svg);
}
catch (DecoderFallbackException)
{
return null;
}
var declared = DeclaredEntity().Matches(text).Select(m => m.Groups[1].Value).ToHashSet(StringComparer.Ordinal);
var result = new StringBuilder(text.Length + 64);
var changed = false;
var position = 0;
foreach (Match skip in Untouched().Matches(text))
{
changed |= Repair(text.AsSpan(position, skip.Index - position), declared, result);
result.Append(skip.Value);
position = skip.Index + skip.Length;
}
changed |= Repair(text.AsSpan(position), declared, result);
if (!changed)
return null;
return Encoding.UTF8.GetBytes(result.ToString()); // a byte order mark was decoded as U+FEFF and stays
}
/// <summary>Appends <paramref name="part"/> to <paramref name="result"/> repaired; true if anything changed.</summary>
private static bool Repair(ReadOnlySpan<char> part, HashSet<string> declared, StringBuilder result)
{
var changed = false;
var text = part.ToString();
var last = 0;
foreach (Match amp in Ampersand().Matches(text))
{
result.Append(text, last, amp.Index - last);
last = amp.Index + amp.Length;
var name = amp.Groups["name"];
if (amp.Groups["number"].Success || name.Success && (XmlEntities.Contains(name.Value) || declared.Contains(name.Value)))
{
result.Append(amp.Value);
continue;
}
changed = true;
if (name.Success && WebUtility.HtmlDecode(amp.Value) is { } decoded && decoded != amp.Value)
{
// An HTML entity: the same character as a reference XML reads.
foreach (var rune in decoded.EnumerateRunes())
result.Append("&#").Append(rune.Value).Append(';');
}
else
{
// A lone "&", or an entity nobody defines: the "&" itself is text.
result.Append("&").Append(amp.Value, 1, amp.Value.Length - 1);
}
}
result.Append(text, last, text.Length - last);
return changed;
}
// "&", then a character reference, a name with its ";", or neither (a lone "&").
[GeneratedRegex(@"&(?:(?<number>#[0-9]+;|#[xX][0-9A-Fa-f]+;)|(?<name>[A-Za-z_][A-Za-z0-9._-]*);)?")]
private static partial Regex Ampersand();
[GeneratedRegex(@"<!--.*?-->|<!\[CDATA\[.*?\]\]>|<!DOCTYPE[^\[>]*(?:\[.*?\])?\s*>", RegexOptions.Singleline)]
private static partial Regex Untouched();
[GeneratedRegex(@"<!ENTITY\s+([A-Za-z_][A-Za-z0-9._-]*)")]
private static partial Regex DeclaredEntity();
}