using System.Globalization; using System.Reflection; using System.Text; using System.Text.RegularExpressions; using UnitsNet; using UnitsNet.Units; namespace NathanMcRae; public class Tsv where T : SaneTsv.TsvRecord { public virtual List Records { get; set; } } public class CommentedTsv : Tsv where T : SaneTsv.TsvRecord { public override List Records { get; set; } public string FileComment { get; set; } = null; } /// /// Sane Tab-Separated Values /// public class SaneTsv { // Like an enum, but more extensible public class ColumnType { } public class StringType : ColumnType { } public class BooleanType : ColumnType { } public class Float32Type : ColumnType { } public class Float32LEType : ColumnType { } public class Float64Type : ColumnType { } public class Float64LEType : ColumnType { } public class UInt32Type : ColumnType { } public class UInt64Type : ColumnType { } public class Int32Type : ColumnType { } public class Int64Type : ColumnType { } public class BinaryType : ColumnType { } public class Iso8601Type : ColumnType { } public class PhysicalUnitsType : ColumnType { public UnitsNet.UnitInfo Units { get; } public ColumnType BaseType { get; internal set; } public PhysicalUnitsType(UnitsNet.UnitInfo units, ColumnType baseType) { Units = units; BaseType = baseType; } } public static readonly string[] ValidUnits = { "m", "s", "A", "K", "cd", "mol", "kg", "Hz", "rad", "sr", "N", "Pa", "J", "W", "C", "V", "F", "Ω", "S", "Wb", "T", "H", "°C", "lm", "lx", "Bq", "Gy", "Sv", "kat" }; public static readonly int MajorVersion = 0; public static readonly int MinorVersion = 0; public static readonly int PatchVersion = 1; public static Regex VersionRegex = new Regex(@"^ ExtraTSV V(\d+)\.(\d+)\.(\d+)"); protected enum FormatType { SIMPLE_TSV = 0, TYPED_TSV = 1, COMMENTED_TSV = 2, } public static readonly byte[] TrueEncoded = Encoding.UTF8.GetBytes("TRUE"); public static readonly byte[] FalseEncoded = Encoding.UTF8.GetBytes("FALSE"); protected static bool? _littleEndian = null; public static bool LittleEndian { get { if (_littleEndian == null) { _littleEndian = BitConverter.GetBytes(double.NegativeInfinity)[7] == 255; } return _littleEndian.Value; } } public static Tsv ParseSimpleTsv(byte[] inputBuffer) where T : TsvRecord, new() { return Parse(inputBuffer, FormatType.SIMPLE_TSV); } public static Tsv ParseTypedTsv(byte[] inputBuffer) where T : TsvRecord, new() { return Parse(inputBuffer, FormatType.TYPED_TSV); } public static CommentedTsv ParseCommentedTsv(byte[] inputBuffer) where T : CommentedTsvRecord, new() { return (CommentedTsv)Parse(inputBuffer, FormatType.COMMENTED_TSV); } public static CommentedTsv ParseExtraTsv(byte[] inputBuffer) where T : CommentedTsvRecord, new() { CommentedTsv parsed = (CommentedTsv)Parse(inputBuffer, FormatType.COMMENTED_TSV); if (parsed.FileComment == null) { throw new Exception($"ExtraTSV expects the file to start with '# ExtraTSV Vx.y.z' where x.y.z is a version compatible with {MajorVersion}.{MinorVersion}.{PatchVersion}"); } Match match = VersionRegex.Match(parsed.FileComment); if (!match.Success) { throw new Exception($"ExtraTSV expects the file to start with '# ExtraTSV Vx.y.z' where x.y.z is a version compatible with {MajorVersion}.{MinorVersion}.{PatchVersion}"); } int fileMajorVersion = int.Parse(match.Groups[1].Value); if (fileMajorVersion != MajorVersion) { throw new Exception($"File has major version ({fileMajorVersion}) which is newer than this parser's version {MajorVersion}"); } return parsed; } protected static Tsv Parse(byte[] inputBuffer, FormatType format) where T : TsvRecord, new() { Tsv parsed; if (format == FormatType.COMMENTED_TSV) { parsed = new CommentedTsv(); } else { parsed = new Tsv(); } parsed.Records = new List(); var columnTypes = new List(); var columnNames = new List(); var columnPropertyInfos = new List(); int columnCount = 0; foreach (PropertyInfo property in typeof(T).GetProperties()) { TsvColumnAttribute attribute = (TsvColumnAttribute)Attribute.GetCustomAttribute(property, typeof(TsvColumnAttribute)); if (attribute == null) { continue; } columnNames.Add(attribute.ColumnName ?? property.Name); columnTypes.Add(attribute.ColumnType ?? GetColumnFromType(property.PropertyType)); columnPropertyInfos.Add(property); // TODO: Check that the property type and given column type are compatible columnCount++; } var fieldBytes = new List(); var fields = new List(); var currentComment = new StringBuilder(); int numFields = -1; int line = 1; int currentLineStart = 0; for (int i = 0; i < inputBuffer.Count(); i++) { if (inputBuffer[i] == '\\') { if (i + 1 == inputBuffer.Count()) { throw new Exception($"Found '\\' at end of input"); } if (inputBuffer[i + 1] == 'n') { fieldBytes.Add((byte)'\n'); i++; } else if (inputBuffer[i + 1] == '\\') { fieldBytes.Add((byte)'\\'); i++; } else if (inputBuffer[i + 1] == 't') { fieldBytes.Add((byte)'\t'); i++; } else if (inputBuffer[i + 1] == '#') { fieldBytes.Add((byte)'#'); i++; } else { throw new Exception($"Expected 'n', 't', '#', or '\\' after '\\' at line {line} column {i - currentLineStart}"); } } else if (inputBuffer[i] == '\t') { // end of field fields.Add(fieldBytes.ToArray()); fieldBytes.Clear(); } else if (inputBuffer[i] == '\n') { fields.Add(fieldBytes.ToArray()); fieldBytes.Clear(); int numTypesBlank = 0; for (int j = 0; j < fields.Count; j++) { string columnString; try { columnString = Encoding.UTF8.GetString(fields[j]); } catch (Exception e) { throw new Exception($"Header field {fields.Count} is not valid UTF-8", e); } string[] columnTypeStrings; string columnName; if (columnString.Contains(':')) { if (format == FormatType.SIMPLE_TSV) { throw new Exception($"Header field {j} contains ':', which is not allowed for column names"); } columnTypeStrings = columnString.Split(":"); columnName = string.Join(":", columnTypeStrings.Take(columnTypeStrings.Length - 1)); } else { if (format > FormatType.SIMPLE_TSV) { throw new Exception($"Header field {fields.Count} has no type"); } columnTypeStrings = new string[] { "" }; columnName = columnString; } ColumnType type; switch (columnTypeStrings.Last()) { case "": numTypesBlank++; type = new StringType(); break; case "string": if (columnTypeStrings.Length > 2 && columnTypeStrings[columnTypeStrings.Length - 2] == "iso8601") { type = new Iso8601Type(); columnName = string.Join(":", columnTypeStrings.Take(columnTypeStrings.Length - 2)); } // TODO: ISO8601 time spans // TODO: ISO8601 time durations else { type = new StringType(); } break; case "boolean": type = new BooleanType(); break; case "float32": type = new Float32Type(); break; case "float32-le": type = new Float32LEType(); break; case "float64": if (columnTypeStrings.Length > 3 && columnTypeStrings[columnTypeStrings.Length - 2] == UnitsTypeText) { string unitName = columnTypeStrings[columnTypeStrings.Length - 3]; if (UnitsNet.Quantity.TryFromUnitAbbreviation(1, unitName, out UnitsNet.IQuantity quantity)) { type = new PhysicalUnitsType(UnitsNet.Quantity.GetUnitInfo(quantity.Unit), new Float64Type()); } else { throw new Exception($"Invalid units: {unitName}"); } columnName = string.Join(":", columnTypeStrings.Take(columnTypeStrings.Length - 3)); } else { type = new Float64Type(); } break; case "float64-le": if (columnTypeStrings.Length > 3 && columnTypeStrings[columnTypeStrings.Length - 2] == UnitsTypeText) { string unitName = columnTypeStrings[columnTypeStrings.Length - 3]; if (UnitsNet.Quantity.TryFromUnitAbbreviation(1, unitName, out UnitsNet.IQuantity quantity)) { type = new PhysicalUnitsType(UnitsNet.Quantity.GetUnitInfo(quantity.Unit), new Float64LEType()); } else { throw new Exception($"Invalid units: {unitName}"); } columnName = string.Join(":", columnTypeStrings.Take(columnTypeStrings.Length - 3)); } else { type = new Float64LEType(); } break; case "uint32": type = new UInt32Type(); break; case "uint64": type = new UInt64Type(); break; case "int32": type = new Int32Type(); break; case "int64": type = new Int64Type(); break; case "binary": type = new BinaryType(); break; default: throw new Exception($"Invalid type '{columnTypeStrings.Last()}' for column {j}"); } // TODO: Allow lax parsing (only worry about parsing columns that are given in the specifying type if (columnNames[j] != columnName) { throw new Exception($"Column {j} has name {columnName}, but expected {columnNames[j]}"); } if (columnTypes[j].GetType() != type.GetType()) { throw new Exception($"Column {j} has type {type}, but expected {columnTypes[j]}"); } } if (currentComment.Length > 0) { if (parsed is CommentedTsv commentedParsed) { commentedParsed.FileComment = currentComment.ToString(); currentComment.Clear(); } else { throw new Exception("Found a file comment, but parser wasn't expecting a comment"); } } fields.Clear(); line++; currentLineStart = i + 1; // Done parsing header break; } else if (inputBuffer[i] == '#') { if (i == currentLineStart && format >= FormatType.COMMENTED_TSV) { int j = i; for (; j < inputBuffer.Length && inputBuffer[j] != '\n'; j++) { } if (j < inputBuffer.Length) { var commentBytes = new byte[j - i - 1]; Array.Copy(inputBuffer, i + 1, commentBytes, 0, j - i - 1); if (currentComment.Length > 0) { currentComment.Append('\n'); } currentComment.Append(Encoding.UTF8.GetString(commentBytes)); i = j; currentLineStart = i + 1; line++; } else { throw new Exception("Comments at end of file are not allowed"); } } else { throw new Exception($"Found unescaped '#' at line {line}, column {i - currentLineStart}"); } } else { fieldBytes.Add(inputBuffer[i]); } } // TODO: need to figure out where the crossover is // Complication: it probably depends on processor count if (inputBuffer.Length < 10000) { parsed.Records.AddRange(Parse(inputBuffer, format, columnPropertyInfos.ToArray(), columnTypes.ToArray(), currentLineStart - 1, inputBuffer.Length)); return parsed; } else { int parseStart = currentLineStart; int tasks = Environment.ProcessorCount - 1; int splitCount = (inputBuffer.Length - parseStart) / tasks; T[][] parsedValues = new T[tasks][]; Parallel.For(0, tasks, i => { int startIndex = i * splitCount + parseStart - 1; int endIndex; if (i == tasks - 1) { endIndex = inputBuffer.Length; } else { endIndex = (i + 1) * splitCount + parseStart; } parsedValues[i] = Parse(inputBuffer, format, columnPropertyInfos.ToArray(), columnTypes.ToArray(), startIndex, endIndex); }); // TODO: Handle relative line numbers for (int i = 0; i < tasks; i++) { parsed.Records.AddRange(parsedValues[i]); } } return parsed; } // This approach is slightly different than others. We skip the record that startIndex is in and // include the record that endIndex is in. We do this because in order to include the record // startIndex is in we'd have to go back to the start of the record's comment, and to know // exactly where that comment started we'd have to go back to the start of the record before that // (not including that other record's comment). protected static T[] Parse(byte[] inputBuffer, FormatType format, PropertyInfo[] columnPropertyInfos, ColumnType[] columnTypes, int startIndex, int endIndex) where T : TsvRecord, new() { var fieldBytes = new List(); var fields = new List(); var currentComment = new StringBuilder(); List parsed = new List(); bool parsingLastRecord = false; int relativeLine = 0; int i = startIndex; while (i < inputBuffer.Length - 1 && inputBuffer[i] != '\n' && inputBuffer[i + 1] != '#') { i++; } if (i >= inputBuffer.Length - 1) { return Array.Empty(); } // Start parsing after \n i++; int currentLineStart = i; for (; i < inputBuffer.Length && (i < endIndex || parsingLastRecord); i++) { if (inputBuffer[i] == '\\') { if (i + 1 == inputBuffer.Count()) { throw new Exception($"Found '\\' at end of input"); } if (inputBuffer[i + 1] == 'n') { fieldBytes.Add((byte)'\n'); i++; } else if (inputBuffer[i + 1] == '\\') { fieldBytes.Add((byte)'\\'); i++; } else if (inputBuffer[i + 1] == 't') { fieldBytes.Add((byte)'\t'); i++; } else if (inputBuffer[i + 1] == '#') { fieldBytes.Add((byte)'#'); i++; } else { throw new Exception($"Expected 'n', 't', '#', or '\\' after '\\' at line {relativeLine} column {i - currentLineStart}"); } } else if (inputBuffer[i] == '\t') { // end of field fields.Add(fieldBytes.ToArray()); fieldBytes.Clear(); } else if (inputBuffer[i] == '\n') { fields.Add(fieldBytes.ToArray()); fieldBytes.Clear(); if (columnTypes.Length != fields.Count) { throw new Exception($"Expected {columnTypes.Length} fields on line {relativeLine}, but found {fields.Count}"); } else { string comment = null; if (currentComment.Length > 0) { comment = currentComment.ToString(); currentComment.Clear(); } parsed.Add(ParseCurrentRecord(columnTypes.ToArray(), columnPropertyInfos.ToArray(), fields, comment, relativeLine)); fields.Clear(); } parsingLastRecord = false; relativeLine++; currentLineStart = i + 1; } else if (inputBuffer[i] == '#') { if (i == currentLineStart && format >= FormatType.COMMENTED_TSV) { int j = i; for (; j < inputBuffer.Length && inputBuffer[j] != '\n'; j++) { } if (j < inputBuffer.Length) { var commentBytes = new byte[j - i - 1]; Array.Copy(inputBuffer, i + 1, commentBytes, 0, j - i - 1); if (currentComment.Length > 0) { currentComment.Append('\n'); } currentComment.Append(Encoding.UTF8.GetString(commentBytes)); i = j; currentLineStart = i + 1; relativeLine++; } else { throw new Exception("Comments at end of file are not allowed"); } } else { throw new Exception($"Found unescaped '#' at line {relativeLine}, column {i - currentLineStart}"); } } else { fieldBytes.Add(inputBuffer[i]); } if (i == endIndex - 1) { parsingLastRecord = true; } } if (endIndex < inputBuffer.Length) { return parsed.ToArray(); } fields.Add(fieldBytes.ToArray()); if (fields.Count == 0) { // TODO throw new Exception("Not sure when this will happen. THis might actuall be fine"); } if (fields.Count != columnTypes.Length) { throw new Exception($"Expected {columnTypes} fields on line {relativeLine}, but found {fields.Count}"); } else { string comment = null; if (currentComment.Length > 0) { comment = currentComment.ToString(); currentComment.Clear(); } parsed.Add(ParseCurrentRecord(columnTypes.ToArray(), columnPropertyInfos.ToArray(), fields, comment, relativeLine)); fields.Clear(); } return parsed.ToArray(); } protected static T ParseCurrentCommentedRecord(ColumnType[] columnTypes, PropertyInfo[] properties, List fields, string comment, int line) where T : CommentedTsvRecord, new() { return (T)ParseCurrentRecord(columnTypes, properties, fields, comment, line); } protected static T ParseCurrentRecord(ColumnType[] columnTypes, PropertyInfo[] properties, List fields, string comment, int line) where T : TsvRecord, new() { T record = new T(); if (record is CommentedTsvRecord commentedRecord) { commentedRecord.Comment = comment; } else if (comment != null) { throw new Exception($"Found comment for line {line}, but format does not support comments"); } record.Line = line; for (int j = 0; j < fields.Count; j++) { // All other types require the content to be UTF-8. Binary fields can ignore that. if (columnTypes[j].GetType() == typeof(BinaryType)) { // TODO: Use faster method for property setting // e.g. https://blog.marcgravell.com/2012/01/playing-with-your-member.html // or https://stackoverflow.com/questions/1027980/improving-performance-reflection-what-alternatives-should-i-consider // or https://stackoverflow.com/questions/12767091/why-are-propertyinfo-setvalue-and-getvalue-so-slow properties[j].SetValue(record, fields[j]); continue; } else if (columnTypes[j].GetType() == typeof(Float32LEType)) { byte[] floatBytes; if (!LittleEndian) { floatBytes = new byte[sizeof(float)]; for (int k = 0; k < sizeof(float); k++) { floatBytes[k] = fields[j][sizeof(float) - 1 - k]; } } else { floatBytes = fields[j]; } properties[j].SetValue(record, BitConverter.ToSingle(floatBytes, 0)); continue; } else if (columnTypes[j].GetType() == typeof(Float64LEType) || (columnTypes[j] is PhysicalUnitsType f64PhUnit && f64PhUnit.BaseType is Float64LEType)) { byte[] floatBytes; if (!LittleEndian) { floatBytes = new byte[sizeof(double)]; for (int k = 0; k < sizeof(double); k++) { floatBytes[k] = fields[j][sizeof(double) - 1 - k]; } } else { floatBytes = fields[j]; } double value = BitConverter.ToDouble(floatBytes, 0); if (columnTypes[j] is PhysicalUnitsType unit) { properties[j].SetValue(record, UnitsNet.Quantity.From(value, unit.Units.Value)); } else { properties[j].SetValue(record, value); } continue; } string fieldString; try { fieldString = Encoding.UTF8.GetString(fields[j]); } catch (Exception e) { throw new Exception($"Field {j} on line {line} is not valid UTF-8", e); } // TODO: Add checking for numeric types format if (columnTypes[j].GetType() == typeof(StringType)) { properties[j].SetValue(record, fieldString); } else if (columnTypes[j].GetType() == typeof(BooleanType)) { bool parsedBool; if (fieldString == "TRUE") { parsedBool = true; } else if (fieldString == "FALSE") { parsedBool = false; } else { throw new Exception($"Field {j} on line {line} is not valid boolean. Must be 'TRUE' or 'FALSE' exactly"); } properties[j].SetValue(record, parsedBool); } else if (columnTypes[j].GetType() == typeof(Float32Type)) { float parsedFloat; if (!float.TryParse(fieldString, out parsedFloat)) { if (fieldString == "-inf") { parsedFloat = float.NegativeInfinity; } else if (fieldString == "+inf") { parsedFloat = float.PositiveInfinity; } else { throw new Exception($"Field {j} on line {line} is not valid single-precision float"); } } properties[j].SetValue(record, parsedFloat); } else if (columnTypes[j].GetType() == typeof(Float64Type) || (columnTypes[j] is PhysicalUnitsType f64PhUnit && f64PhUnit.BaseType is Float64Type)) { double parsedDouble; if (!double.TryParse(fieldString, out parsedDouble)) { if (fieldString == "-inf") { parsedDouble = float.NegativeInfinity; } else if (fieldString == "+inf") { parsedDouble = float.PositiveInfinity; } else { throw new Exception($"Field {j} on line {line} is not valid double-precision float"); } } if (columnTypes[j] is PhysicalUnitsType unit) { properties[j].SetValue(record, UnitsNet.Quantity.From(parsedDouble, unit.Units.Value)); } else { properties[j].SetValue(record, parsedDouble); } } else if (columnTypes[j].GetType() == typeof(UInt32Type)) { if (!UInt32.TryParse(fieldString, out UInt32 parsedUInt32)) { throw new Exception($"Field {j} on line {line} is not valid UInt32"); } properties[j].SetValue(record, parsedUInt32); } else if (columnTypes[j].GetType() == typeof(UInt64Type)) { if (!UInt64.TryParse(fieldString, out UInt64 parsedUInt64)) { throw new Exception($"Field {j} on line {line} is not valid UInt64"); } properties[j].SetValue(record, parsedUInt64); } else if (columnTypes[j].GetType() == typeof(Int32Type)) { if (!Int32.TryParse(fieldString, out Int32 parsedInt32)) { throw new Exception($"Field {j} on line {line} is not valid Int32"); } properties[j].SetValue(record, parsedInt32); } else if (columnTypes[j].GetType() == typeof(Int64Type)) { if (!Int64.TryParse(fieldString, out Int64 parsedInt64)) { throw new Exception($"Field {j} on line {line} is not valid Int64"); } properties[j].SetValue(record, parsedInt64); } else if (columnTypes[j].GetType() == typeof(Iso8601Type)) { if (!DateTime.TryParseExact(fieldString, "yyyy-MM-ddTHH:mm:ss.ffff", CultureInfo.InvariantCulture, DateTimeStyles.None, out DateTime parsed)) { throw new Exception($"ISO 8601 timestamp format error on line {line}, field {j}"); } properties[j].SetValue(record, parsed); } else { throw new Exception($"Unexpected type {columnTypes[j]}"); } } return record; } public static byte[] SerializeSimpleTsv(IList header, IList> data) { var serialized = new List(); var escapedString = new StringBuilder(); // Serialize header for (int i = 0; i < header.Count; i++) { if (header[i].Contains(':')) { throw new Exception($"Column {i} contains the character ':'"); } for (int j = i + 1; j < header.Count; j++) { if (header[i] == header[j]) { throw new Exception("Column names in header must be unique"); } } for (int j = 0; j < header[i].Count(); j++) { if (header[i][j] == '\n') { escapedString.Append("\\n"); } else if (header[i][j] == '\t') { escapedString.Append("\\t"); } else if (header[i][j] == '\\') { escapedString.Append("\\\\"); } else if (header[i][j] == '#') { escapedString.Append("\\#"); } else { escapedString.Append(header[i][j]); } } if (i == header.Count - 1) { escapedString.Append('\n'); } else { escapedString.Append('\t'); } } serialized.AddRange(Encoding.UTF8.GetBytes(escapedString.ToString())); // TODO: need to figure out where the crossover it // Complication: it probably depends on processor count if (data.Count < 100) { serialized.AddRange(Encoding.UTF8.GetBytes(SerializeSimpleTsv(data, 0, data.Count))); } else { int tasks = Environment.ProcessorCount - 1; int splitCount = data.Count / tasks; byte[][] bytes = new byte[tasks][]; Parallel.For(0, tasks, i => { int endIndex; if (i == tasks - 1) { endIndex = data.Count; } else { endIndex = (i + 1) * splitCount; } string escapedString = SerializeSimpleTsv(data, i * splitCount, endIndex); bytes[i] = Encoding.UTF8.GetBytes(escapedString); }); for (int i = 0; i < tasks; i++) { serialized.AddRange(bytes[i]); } } return serialized.ToArray(); } public static string SerializeSimpleTsv(IList> data, int startIndex, int endIndex) { var escapedString = new StringBuilder(); // Serialize data for (int i = startIndex; i < endIndex; i++) { for (int j = 0; j < data[i].Count; j++) { for (int k = 0; k < data[i][j].Length; k++) { if (data[i][j][k] == '\n') { escapedString.Append("\\n"); } else if (data[i][j][k] == '\t') { escapedString.Append("\\t"); } else if (data[i][j][k] == '\\') { escapedString.Append("\\\\"); } else if (data[i][j][k] == '#') { escapedString.Append("\\#"); } else { escapedString.Append(data[i][j][k]); } } if (j < data[i].Count - 1) { escapedString.Append('\t'); } else if (i < data.Count - 1) { escapedString.Append('\n'); } } } return escapedString.ToString(); } public static (string[] columns, string[][] data) ParseSimpleTsv(byte[] inputBuffer) { string[] columnNames = null; var headerFields = new List(); var fieldBytes = new List(); int startOfData = -1; for (int i = 0; i < inputBuffer.Count(); i++) { if (inputBuffer[i] == '\\') { if (i + 1 == inputBuffer.Count()) { throw new Exception($"Found '\\' at end of input"); } if (inputBuffer[i + 1] == 'n') { fieldBytes.Add((byte)'\n'); i++; } else if (inputBuffer[i + 1] == '\\') { fieldBytes.Add((byte)'\\'); i++; } else if (inputBuffer[i + 1] == 't') { fieldBytes.Add((byte)'\t'); i++; } else if (inputBuffer[i + 1] == '#') { fieldBytes.Add((byte)'#'); i++; } else { throw new Exception($"Expected 'n', 't', '#', or '\\' after '\\' at line {1} column {i}"); } } else if (inputBuffer[i] == '\t') { // end of field headerFields.Add(fieldBytes.ToArray()); fieldBytes.Clear(); } else if (inputBuffer[i] == '\n') { // This is the end of the header headerFields.Add(fieldBytes.ToArray()); startOfData = i + 1; columnNames = new string[headerFields.Count]; fieldBytes.Clear(); for (int j = 0; j < headerFields.Count; j++) { string columnString; try { columnString = Encoding.UTF8.GetString(headerFields[j]); } catch (Exception e) { throw new Exception($"Column {headerFields.Count} name is not valid UTF-8", e); } if (columnString.Contains(':')) { throw new Exception($"Header field {headerFields.Count} contain ':', which is not allowed for column names"); } columnNames[j] = columnString; } // Done parsing header break; } else if (inputBuffer[i] == '#') { throw new Exception($"Found unescaped '#' at line 1, column {i}"); } else { fieldBytes.Add(inputBuffer[i]); } } return (columnNames, ParseSimpleTsv(inputBuffer, columnNames.Length, startOfData, inputBuffer.Length)); } public static string[][] ParseSimpleTsv(byte[] inputBuffer, int numFields, int startIndex, int endIndex) { var fieldBytes = new List(); var fields = new List(); var records = new List(); int line = 2; int currentLineStart = 0; // Go back to the start of the current line int i = startIndex; while (inputBuffer[i] != '\n') { i--; } // We want to start at the first byte of the current line i++; for (; i < endIndex; i++) { if (inputBuffer[i] == '\\') { if (i + 1 == inputBuffer.Count()) { throw new Exception($"Found '\\' at end of input"); } if (inputBuffer[i + 1] == 'n') { fieldBytes.Add((byte)'\n'); i++; } else if (inputBuffer[i + 1] == '\\') { fieldBytes.Add((byte)'\\'); i++; } else if (inputBuffer[i + 1] == 't') { fieldBytes.Add((byte)'\t'); i++; } else if (inputBuffer[i + 1] == '#') { fieldBytes.Add((byte)'#'); i++; } else { throw new Exception($"Expected 'n', 't', '#', or '\\' after '\\' at line {line} column {i - currentLineStart}"); } } else if (inputBuffer[i] == '\t') { // end of field fields.Add(fieldBytes.ToArray()); fieldBytes.Clear(); } else if (inputBuffer[i] == '\n') { fields.Add(fieldBytes.ToArray()); fieldBytes.Clear(); if (numFields != fields.Count) { throw new Exception($"Expected {numFields} fields on line {line}, but found {fields.Count}"); } else { var fieldStrings = new string[fields.Count]; for (int j = 0; j < fields.Count; j++) { try { fieldStrings[j] = Encoding.UTF8.GetString(fields[j]); } catch (Exception e) { throw new Exception($"Line {line}, column {j} is not valid UTF-8", e); } } records.Add(fieldStrings); fields.Clear(); } line++; currentLineStart = i + 1; } else if (inputBuffer[i] == '#') { throw new Exception($"Found unescaped '#' at line {line}, column {i - currentLineStart}"); } else { fieldBytes.Add(inputBuffer[i]); } } fields.Add(fieldBytes.ToArray()); if (fields.Count == 0 && endIndex == inputBuffer.Length) { throw new Exception("Found 0 fields on last line. Possibly because of extra \\n after last record"); } if (numFields != fields.Count) { if (endIndex == inputBuffer.Length) { throw new Exception($"Expected {numFields} fields on line {line}, but found {fields.Count}"); } else { return records.ToArray(); } } else { var fieldStrings = new string[fields.Count]; for (int j = 0; j < fields.Count; j++) { try { fieldStrings[j] = Encoding.UTF8.GetString(fields[j]); } catch (Exception e) { throw new Exception($"Line {line}, column {j} is not valid UTF-8", e); } } records.Add(fieldStrings); fields.Clear(); } return records.ToArray(); } public static string UnitsTypeText = "ph-unit"; public static Regex UnitsRegex = new Regex("([^:]+):" + UnitsTypeText + ":(float32|float32-le|float64|float64-le|uint32|uint64|int32|int64)"); public static ColumnType GetColumnFromString(string type) { if (type == "string") { return new StringType(); } else if (type == "boolean") { return new BooleanType(); } else if (type == "float32") { return new Float32Type(); } else if (type == "float32-le") { return new Float32LEType(); } else if (type == "float64") { return new Float64Type(); } else if (type == "float64-le") { return new Float64LEType(); } else if (type == "uint32") { return new UInt32Type(); } else if (type == "uint64") { return new UInt64Type(); } else if (type == "int32") { return new Int32Type(); } else if (type == "int64") { return new Int64Type(); } else if (type == "binary") { return new BinaryType(); } else if (type == "iso8601") { return new Iso8601Type(); } else if (UnitsRegex.IsMatch(type)) { Match match = UnitsRegex.Match(type); string unitName = match.Groups[1].Value; string baseType = match.Groups[2].Value; return new PhysicalUnitsType(ParseUnit(unitName), GetColumnFromString(baseType)); //if (UnitsNet.Quantity.TryFromUnitAbbreviation(1, unitName, out UnitsNet.IQuantity quantity)) //{ // return new PhysicalUnitsType(UnitsNet.Quantity.GetUnitInfo(quantity.Unit), GetColumnFromString(baseType)); //} //else //{ // throw new Exception($"Invalid units: {unitName}"); //} } else { throw new Exception($"Invalid type: {type.GetType()}"); } } public static ColumnType GetColumnFromType(Type type) { if (type == typeof(string)) { return new StringType(); } else if (type == typeof(bool)) { return new BooleanType(); } else if (type == typeof(float)) { return new Float32Type(); } else if (type == typeof(double)) { return new Float64Type(); } else if (type == typeof(UInt32)) { return new UInt32Type(); } else if (type == typeof(UInt64)) { return new UInt64Type(); } else if (type == typeof(Int32)) { return new Int32Type(); } else if (type == typeof(Int64)) { return new Int64Type(); } else if (type == typeof(byte[])) { return new BinaryType(); } else if (type == typeof(DateTime)) { return new Iso8601Type(); } else if (type == typeof(UnitsNet.Mass)) { // TODO //UnitsNet.UnitInfo a = new UnitsNet.UnitInfo([d]) var a = new UnitsNet.UnitInfo(UnitsNet.Units.MassUnit.Kilogram, "kgs", new UnitsNet.BaseUnits(mass: UnitsNet.Units.MassUnit.Kilogram)); return new PhysicalUnitsType(a, new Float64Type()); } else { throw new Exception($"Invalid type: {type.GetType()}"); } } public static string GetNameFromColumn(ColumnType type) { if (type.GetType() == typeof(StringType)) { return "string"; } else if (type.GetType() == typeof(BooleanType)) { return "boolean"; } else if (type.GetType() == typeof(Float32Type)) { return "float32"; } else if (type.GetType() == typeof(Float32LEType)) { return "float32-le"; } else if (type.GetType() == typeof(Float64Type)) { return "float64"; } else if (type.GetType() == typeof(Float64LEType)) { return "float64-le"; } else if (type.GetType() == typeof(UInt32Type)) { return "uint32"; } else if (type.GetType() == typeof(UInt64Type)) { return "uint64"; } else if (type.GetType() == typeof(Int32Type)) { return "int32"; } else if (type.GetType() == typeof(Int64Type)) { return "int64"; } else if (type.GetType() == typeof(BinaryType)) { return "binary"; } else if (type.GetType() == typeof(Iso8601Type)) { return "iso8601:string"; } else if (type is PhysicalUnitsType unit) { return $"{unit.Units.Name}:{UnitsTypeText}:{GetNameFromColumn(unit.BaseType)}"; } else { throw new Exception($"Invalid type: {type.GetType()}"); } } public static byte[] SerializeSimpleTsv(IList data) where T : TsvRecord { return SerializeTsv(data, FormatType.SIMPLE_TSV); } public static byte[] SerializeTypedTsv(IList data) where T : TsvRecord { return SerializeTsv(data, FormatType.TYPED_TSV); } public static byte[] SerializeCommentedTsv(IList data, string fileComment) where T : CommentedTsvRecord { return SerializeTsv(data, FormatType.COMMENTED_TSV, fileComment); } public static byte[] SerializeExtraTsv(IList data) where T : TsvRecord { return SerializeTsv(data, FormatType.COMMENTED_TSV, $" ExtraTSV V{MajorVersion}.{MinorVersion}.{PatchVersion}"); } protected static byte[] SerializeTsv(IList data, FormatType tsvFormat, string fileComment = null) { var bytes = new List(); if (fileComment != null) { if (tsvFormat != FormatType.COMMENTED_TSV) { throw new Exception($"File comments are not valid for {tsvFormat}"); } bytes.AddRange(Encoding.UTF8.GetBytes("#" + fileComment.Replace("\n", "\n#") + "\n")); } var columnTypes = new List(); var columnNames = new List(); var columnPropertyInfos = new List(); int columnCount = 0; // Serialize header foreach (PropertyInfo property in typeof(T).GetProperties()) { TsvColumnAttribute attribute = (TsvColumnAttribute)Attribute.GetCustomAttribute(property, typeof(TsvColumnAttribute)); if (attribute == null) { continue; } string headerName = attribute.ColumnName ?? property.Name; columnNames.Add(headerName); ColumnType headerType = attribute.ColumnType ?? GetColumnFromType(property.PropertyType); if (tsvFormat == FormatType.SIMPLE_TSV && headerType.GetType() != typeof(StringType)) { throw new Exception($"Serializing Simple TSV requires all columns be of type string, but column '{headerName}' has type '{headerType}'"); } columnTypes.Add(headerType); columnPropertyInfos.Add(property); // TODO: Check that the property type and given column type are compatible columnCount++; } // Serialize header for (int i = 0; i < columnNames.Count; i++) { for (int j = i + 1; j < columnNames.Count; j++) { if (columnNames[i] == columnNames[j]) { throw new Exception("Column names in header must be unique"); } } byte[] nameEncoded = Encoding.UTF8.GetBytes(columnNames[i]); for (int j = 0; j < nameEncoded.Length; j++) { if (nameEncoded[j] == '\n') { bytes.Add((byte)'\\'); bytes.Add((byte)'n'); } else if (nameEncoded[j] == '\t') { bytes.Add((byte)'\\'); bytes.Add((byte)'t'); } else if (nameEncoded[j] == '\\') { bytes.Add((byte)'\\'); bytes.Add((byte)'\\'); } else if (nameEncoded[j] == '#') { bytes.Add((byte)'\\'); bytes.Add((byte)'#'); } else { bytes.Add(nameEncoded[j]); } } if (tsvFormat != FormatType.SIMPLE_TSV) { bytes.Add((byte)':'); try { bytes.AddRange(Encoding.UTF8.GetBytes(GetNameFromColumn(columnTypes[i]))); } catch (Exception e) { throw new Exception($"Invalid column type for column {i}", e); } } if (i == columnNames.Count - 1) { bytes.Add((byte)'\n'); } else { bytes.Add((byte)'\t'); } } // Serialize data SerializeTsv(data, bytes, columnPropertyInfos.ToArray(), columnTypes.ToArray(), tsvFormat, 0, data.Count); return bytes.ToArray(); } protected static void SerializeTsv(IList data, List bytes, PropertyInfo[] columnPropertyInfos, ColumnType[] columnTypes, FormatType tsvFormat, int startIndex, int endIndex) { // Serialize data for (int i = 0; i < data.Count; i++) { for (int j = 0; j < columnTypes.Length; j++) { object datum = columnPropertyInfos[j].GetValue(data[i]); try { byte[] fieldEncoded = null; // Some fields definitely don't need escaping, so we add them directly to bytes bool skipEscaping = false; if (columnTypes[j].GetType() == typeof(StringType)) { fieldEncoded = Encoding.UTF8.GetBytes((string)datum); } else if (columnTypes[j].GetType() == typeof(BooleanType)) { bytes.AddRange((bool)datum ? TrueEncoded : FalseEncoded); skipEscaping = true; } else if (columnTypes[j].GetType() == typeof(Float32Type)) { if (datum is float f) { if (float.IsNegativeInfinity(f)) { bytes.AddRange(Encoding.UTF8.GetBytes("-inf")); } else if (float.IsPositiveInfinity(f)) { bytes.AddRange(Encoding.UTF8.GetBytes("+inf")); } else { // See https://learn.microsoft.com/en-us/dotnet/standard/base-types/standard-numeric-format-strings#round-trip-format-specifier-r bytes.AddRange(Encoding.UTF8.GetBytes(((float)datum).ToString("G9"))); } } else { throw new InvalidCastException(); } skipEscaping = true; } else if (columnTypes[j].GetType() == typeof(Float32LEType)) { if (LittleEndian) { fieldEncoded = BitConverter.GetBytes((float)datum); } else { byte[] floatBytes = BitConverter.GetBytes((float)datum); fieldEncoded = new byte[sizeof(float)]; for (int k = 0; k < sizeof(float); k++) { fieldEncoded[k] = floatBytes[sizeof(float) - 1 - k]; } } } else if (columnTypes[j].GetType() == typeof(Float64Type) || (columnTypes[j] is PhysicalUnitsType f64PhUnit && f64PhUnit.BaseType is Float64Type)) { double value; if (datum is double d) { value = d; } // TODO: check units match else if (datum is UnitsNet.IQuantity quantity) { value = quantity.Value; } else { throw new InvalidCastException(); } if (double.IsNegativeInfinity(value)) { bytes.AddRange(Encoding.UTF8.GetBytes("-inf")); } else if (double.IsPositiveInfinity(value)) { bytes.AddRange(Encoding.UTF8.GetBytes("+inf")); } else { // See https://learn.microsoft.com/en-us/dotnet/standard/base-types/standard-numeric-format-strings#round-trip-format-specifier-r bytes.AddRange(Encoding.UTF8.GetBytes((value).ToString("G17"))); } skipEscaping = true; } else if (columnTypes[j].GetType() == typeof(Float64LEType) || (columnTypes[j] is PhysicalUnitsType f64LEPhUnit && f64LEPhUnit.BaseType is Float64LEType)) { double value; if (datum is double d) { value = d; } // TODO: check units match else if (datum is UnitsNet.IQuantity quantity) { value = quantity.Value; } else { throw new InvalidCastException(); } if (LittleEndian) { fieldEncoded = BitConverter.GetBytes((double)value); } else { byte[] doubleBytes = BitConverter.GetBytes((double)value); fieldEncoded = new byte[sizeof(double)]; for (int k = 0; k < sizeof(double); k++) { fieldEncoded[k] = doubleBytes[sizeof(double) - 1 - k]; } } } else if (columnTypes[j].GetType() == typeof(UInt32Type)) { bytes.AddRange(Encoding.UTF8.GetBytes(((UInt32)datum).ToString())); skipEscaping = true; } else if (columnTypes[j].GetType() == typeof(UInt64Type)) { bytes.AddRange(Encoding.UTF8.GetBytes(((UInt64)datum).ToString())); skipEscaping = true; } else if (columnTypes[j].GetType() == typeof(Int32Type)) { bytes.AddRange(Encoding.UTF8.GetBytes(((Int32)datum).ToString())); skipEscaping = true; } else if (columnTypes[j].GetType() == typeof(Int64Type)) { bytes.AddRange(Encoding.UTF8.GetBytes(((Int64)datum).ToString())); skipEscaping = true; } else if (columnTypes[j].GetType() == typeof(BinaryType)) { fieldEncoded = (byte[])datum; } else if (columnTypes[j].GetType() == typeof(Iso8601Type)) { fieldEncoded = Encoding.UTF8.GetBytes(((DateTime)datum).ToString("yyyy-MM-ddTHH:mm:ss.ffff")); } else if (columnTypes[j] is PhysicalUnitsType phUnits) { throw new NotImplementedException($"Physical units types don't support {GetNameFromColumn(phUnits.BaseType)} as a base type"); } else { throw new Exception($"Unexpected column type {columnTypes[j]} for column {j}"); } if (!skipEscaping) { for (int k = 0; k < fieldEncoded.Length; k++) { if (fieldEncoded[k] == '\n') { bytes.Add((byte)'\\'); bytes.Add((byte)'n'); } else if (fieldEncoded[k] == '\t') { bytes.Add((byte)'\\'); bytes.Add((byte)'t'); } else if (fieldEncoded[k] == '\\') { bytes.Add((byte)'\\'); bytes.Add((byte)'\\'); } else if (fieldEncoded[k] == '#') { bytes.Add((byte)'\\'); bytes.Add((byte)'#'); } else { bytes.Add(fieldEncoded[k]); } } } if (j < columnTypes.Length - 1) { bytes.Add((byte)'\t'); } else if (i < data.Count - 1) { bytes.Add((byte)'\n'); } } catch (InvalidCastException e) { throw new Exception($"Record {i}, field {j} expected type compatible with {GetNameFromColumn(columnTypes[j])}", e); } } } } public class SimpleTsvRecord { public string[] ColumnNames { get; } public string Comment { get; } public string[] Fields { get; } public int? Line { get; } public string this[string columnName] => Fields[Array.IndexOf(ColumnNames, columnName)]; public string this[int columnIndex] => Fields[columnIndex]; public SimpleTsvRecord(string[] columnNames, string[] fields, string comment, int line) { ColumnNames = columnNames; Fields = fields; Comment = comment; Line = line; } } public class TsvRecord { public int? Line { get; set; } public TsvRecord(int? line) { Line = line; } public TsvRecord() { } } public class CommentedTsvRecord : TsvRecord { public string Comment { get; set; } public CommentedTsvRecord(string comment, int? line) { Comment = comment; Line = line; } public CommentedTsvRecord() { } } public class TestRecord : CommentedTsvRecord { [TypedTsvColumn("my-column")] public string MyColumn { get; set; } } // TODO: Add column ordering public class TsvColumnAttribute : Attribute { public string ColumnName { get; } public virtual ColumnType ColumnType { get; } public TsvColumnAttribute() { ColumnType = new StringType(); } public TsvColumnAttribute(string columnName) { ColumnType = new StringType(); ColumnName = columnName; } } // TODO: Add column ordering public class TypedTsvColumnAttribute : TsvColumnAttribute { public override ColumnType ColumnType { get; } public TypedTsvColumnAttribute() { } public TypedTsvColumnAttribute(string columnName) : base(columnName) { } public TypedTsvColumnAttribute(string columnName, string columnType) : base(columnName) { ColumnType = GetColumnFromString(columnType); } public TypedTsvColumnAttribute(ColumnType columnType) { ColumnType = columnType; } } public static UnitInfo ParseUnit(string unitName) { // Find all unit enum types in the UnitsNet namespace var unitEnumTypes = Assembly.GetAssembly(typeof(LengthUnit)) .GetTypes() .Where(t => t.IsEnum && t.Namespace == typeof(LengthUnit).Namespace); foreach (var unitEnumType in unitEnumTypes) { if (UnitParser.Default.TryParse(unitName, unitEnumType, out Enum unitEnum)) { // Successfully parsed the abbreviation, retrieve UnitInfo return Quantity.GetUnitInfo(unitEnum); } } throw new ArgumentException($"Unable to parse unit abbreviation: {unitName}"); } }