nathanmcrae/sane-tsv
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Compare commits

base: nathanmcrae:78eaa5dbabc4cfe1ccb52f44615458743acb182f
Branches Tags
nathanmcrae:master
..
compare: nathanmcrae:f98a40a173fb773d42d48cffc54620f711a5b2c8
Branches Tags
nathanmcrae:master

2 Commits
78eaa5dbab ... f98a40a173

Author SHA1 Message Date
Nathan McRae
f98a40a173 Make parallel implementations the default 2024-03-08 12:31:40 -08:00
Nathan McRae
0c61128e0e Complete parallel parsing implementation 2024-03-08 11:43:06 -08:00
2 changed files with 47 additions and 1031 deletions
Show Stats Expand all files Collapse all files

862
SaneTsv/SaneTsv.cs
View File

@ -75,7 +75,6 @@ public class SaneTsv
return (CommentedTsv<T>)Parse<T>(inputBuffer, FormatType.COMMENTED_TSV); return (CommentedTsv<T>)Parse<T>(inputBuffer, FormatType.COMMENTED_TSV);
} }
// TODO: Have parsing errors include line / column #
protected static Tsv<T> Parse<T>(byte[] inputBuffer, FormatType format) where T : TsvRecord, new() protected static Tsv<T> Parse<T>(byte[] inputBuffer, FormatType format) where T : TsvRecord, new()
{ {
Tsv<T> parsed; Tsv<T> parsed;
@ -113,285 +112,6 @@ public class SaneTsv
var fields = new List<byte[]>(); var fields = new List<byte[]>();
var currentComment = new StringBuilder(); 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();
if (numFields < 0)
{
// This is the header
numFields = fields.Count;
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($"Header {fields.Count} is not valid UTF-8", e);
}
string columnTypeString;
string columnName;
if (columnString.Contains(':'))
{
if (format == FormatType.SIMPLE_TSV)
{
throw new Exception($"Header {fields.Count} contain ':', which is not allowed for column names");
}
columnTypeString = columnString.Split(":").Last();
columnName = columnString.Substring(0, columnString.Length - columnTypeString.Length - 1);
}
else
{
if (format > FormatType.SIMPLE_TSV)
{
throw new Exception($"Header {fields.Count} has no type");
}
columnTypeString = "";
columnName = columnString;
}
Type type;
switch (columnTypeString)
{
case "":
numTypesBlank++;
type = typeof(StringType);
break;
case "string":
type = typeof(StringType);
break;
case "boolean":
type = typeof(BooleanType);
break;
case "float32":
type = typeof(Float32Type);
break;
case "float32-le":
type = typeof(Float32LEType);
break;
case "float64":
type = typeof(Float64Type);
break;
case "float64-le":
type = typeof(Float64LEType);
break;
case "uint32":
type = typeof(UInt32Type);
break;
case "uint64":
type = typeof(UInt64Type);
break;
case "int32":
type = typeof(Int32Type);
break;
case "int64":
type = typeof(Int64Type);
break;
case "binary":
type = typeof(BinaryType);
break;
default:
throw new Exception($"Invalid type '{columnTypeString}' for column {j}");
}
// TODO: Check column name uniqueness
// TODO: Allow lax parsing (only worry about parsing columns that are given in the specifying type
if (headerNames[j] != columnName)
{
throw new Exception($"Column {j} has name {columnName}, but expected {headerNames[j]}");
}
if (headerTypes[j] != type)
{
throw new Exception($"Column {j} has type {type}, but expected {headerTypes[j]}");
}
}
if (currentComment.Length > 0)
{
if (parsed is CommentedTsv<T> commentedParsed)
{
commentedParsed.FileComment = currentComment.ToString();
currentComment.Clear();
}
else
{
throw new Exception("Found a file comment, but parser wasn't expecting a comment");
}
}
fields.Clear();
}
else if (numFields != fields.Count)
{
throw new Exception($"Expected {numFields} fields on line {line}, but found {fields.Count}");
}
else
{
string comment = null;
if (currentComment.Length > 0)
{
comment = currentComment.ToString();
currentComment.Clear();
}
parsed.Records.Add(ParseCurrentRecord<T>(headerTypes.ToArray(), headerPropertyInfos.ToArray(), fields, comment, line));
fields.Clear();
}
line++;
currentLineStart = i + 1;
}
else if (inputBuffer[i] == '#')
{
if (i == currentLineStart && format >= FormatType.COMMENTED_TSV)
{
int j = i;
for (; j < inputBuffer.Length && inputBuffer[j] != '\n'; j++) { }
if (j < inputBuffer.Length)
{
var commentBytes = new byte[j - i - 1];
Array.Copy(inputBuffer, i + 1, commentBytes, 0, j - i - 1);
if (currentComment.Length > 0)
{
currentComment.Append('\n');
}
currentComment.Append(Encoding.UTF8.GetString(commentBytes));
i = j;
currentLineStart = i + 1;
line++;
}
else
{
throw new Exception("Comments at end of file are not allowed");
}
}
else
{
throw new Exception($"Found unescaped '#' at line {line}, column {i - currentLineStart}");
}
}
else
{
fieldBytes.Add(inputBuffer[i]);
}
}
fields.Add(fieldBytes.ToArray());
if (fields.Count == 0)
{
throw new Exception("Found 0 fields on last line. Possibly because of extra \\n after last record");
}
if (numFields != fields.Count)
{
throw new Exception($"Expected {numFields} fields on line {parsed.Records.Count + 2}, but found {fields.Count}");
}
else
{
string comment = null;
if (currentComment.Length > 0)
{
comment = currentComment.ToString();
currentComment.Clear();
}
parsed.Records.Add(ParseCurrentRecord<T>(headerTypes.ToArray(), headerPropertyInfos.ToArray(), fields, comment, line));
fields.Clear();
}
return parsed;
}
protected static Tsv<T> ParseParallel<T>(byte[] inputBuffer, FormatType format) where T : TsvRecord, new()
{
Tsv<T> parsed;
if (format == FormatType.COMMENTED_TSV)
{
parsed = new CommentedTsv<T>();
}
else
{
parsed = new Tsv<T>();
}
parsed.Records = new List<T>();
var headerTypes = new List<Type>();
var headerNames = new List<string>();
var headerPropertyInfos = new List<PropertyInfo>();
int columnCount = 0;
foreach (PropertyInfo property in typeof(T).GetProperties())
{
TypedTsvColumnAttribute attribute = (TypedTsvColumnAttribute)Attribute.GetCustomAttribute(property, typeof(TypedTsvColumnAttribute));
if (attribute == null)
{
continue;
}
headerNames.Add(attribute.ColumnName ?? property.Name);
headerTypes.Add(attribute.ColumnType ?? GetColumnFromType(property.PropertyType));
headerPropertyInfos.Add(property);
// TODO: Check that the property type and given column type are compatible
columnCount++;
}
var fieldBytes = new List<byte>();
var fields = new List<byte[]>();
var currentComment = new StringBuilder();
int numFields = -1; int numFields = -1;
int line = 1; int line = 1;
int currentLineStart = 0; int currentLineStart = 0;
@ -588,9 +308,40 @@ public class SaneTsv
} }
} }
T[] parsedRecords = ParseParallel<T>(inputBuffer, format, headerPropertyInfos.ToArray(), headerTypes.ToArray(), currentLineStart - 1, inputBuffer.Length); // TODO: need to figure out where the crossover is
// Complication: it probably depends on processor count
if (inputBuffer.Length < 10000)
{
parsed.Records.AddRange(Parse<T>(inputBuffer, format, headerPropertyInfos.ToArray(), headerTypes.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;
}
parsed.Records.AddRange(parsedRecords); parsedValues[i] = Parse<T>(inputBuffer, format, headerPropertyInfos.ToArray(), headerTypes.ToArray(), startIndex, endIndex);
});
for (int i = 0; i < tasks; i++)
{
parsed.Records.AddRange(parsedValues[i]);
}
}
return parsed; return parsed;
} }
@ -600,7 +351,7 @@ public class SaneTsv
// startIndex is in we'd have to go back to the start of the record's comment, and to know // 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 // 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). // (not including that other record's comment).
protected static T[] ParseParallel<T>(byte[] inputBuffer, FormatType format, PropertyInfo[] headerPropertyInfos, Type[] headerTypes, int startIndex, int endIndex) where T : TsvRecord, new() protected static T[] Parse<T>(byte[] inputBuffer, FormatType format, PropertyInfo[] headerPropertyInfos, Type[] headerTypes, int startIndex, int endIndex) where T : TsvRecord, new()
{ {
var fieldBytes = new List<byte>(); var fieldBytes = new List<byte>();
var fields = new List<byte[]>(); var fields = new List<byte[]>();
@ -730,6 +481,11 @@ public class SaneTsv
} }
} }
if (endIndex < inputBuffer.Length)
{
return parsed.ToArray();
}
fields.Add(fieldBytes.ToArray()); fields.Add(fieldBytes.ToArray());
if (fields.Count == 0) if (fields.Count == 0)
@ -949,103 +705,6 @@ public class SaneTsv
} }
public static byte[] SerializeSimpleTsv(IList<string> header, IList<IList<string>> data) public static byte[] SerializeSimpleTsv(IList<string> header, IList<IList<string>> data)
{
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');
}
}
// Serialize data
for (int i = 0; i < data.Count; 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 Encoding.UTF8.GetBytes(escapedString.ToString());
}
public static byte[] SerializeSimpleTsvParallel(IList<string> header, IList<IList<string>> data)
{ {
var serialized = new List<byte>(); var serialized = new List<byte>();
var escapedString = new StringBuilder(); var escapedString = new StringBuilder();
@ -1106,7 +765,7 @@ public class SaneTsv
// Complication: it probably depends on processor count // Complication: it probably depends on processor count
if (data.Count < 100) if (data.Count < 100)
{ {
serialized.AddRange(Encoding.UTF8.GetBytes(SerializeSimpleTsvParallel(data, 0, data.Count))); serialized.AddRange(Encoding.UTF8.GetBytes(SerializeSimpleTsv(data, 0, data.Count)));
} }
else else
{ {
@ -1124,7 +783,7 @@ public class SaneTsv
{ {
endIndex = (i + 1) * splitCount; endIndex = (i + 1) * splitCount;
} }
string escapedString = SerializeSimpleTsvParallel(data, i * splitCount, endIndex); string escapedString = SerializeSimpleTsv(data, i * splitCount, endIndex);
bytes[i] = Encoding.UTF8.GetBytes(escapedString); bytes[i] = Encoding.UTF8.GetBytes(escapedString);
}); });
@ -1137,7 +796,7 @@ public class SaneTsv
return serialized.ToArray(); return serialized.ToArray();
} }
public static string SerializeSimpleTsvParallel(IList<IList<string>> data, int startIndex, int endIndex) public static string SerializeSimpleTsv(IList<IList<string>> data, int startIndex, int endIndex)
{ {
var escapedString = new StringBuilder(); var escapedString = new StringBuilder();
@ -1184,159 +843,7 @@ public class SaneTsv
return escapedString.ToString(); return escapedString.ToString();
} }
public static (string[] columns, string[][] data) ParseSimpleTsv(byte[] inputBuffer) public static (string[] columns, string[][] data) ParseSimpleTsv(byte[] inputBuffer)
{
string[] columnNames = null;
var fieldBytes = new List<byte>();
var fields = new List<byte[]>();
var records = new List<string[]>();
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();
if (numFields < 0)
{
// This is the header
numFields = fields.Count;
columnNames = new string[numFields];
for (int j = 0; j < fields.Count; j++)
{
string columnString;
try
{
columnString = Encoding.UTF8.GetString(fields[j]);
}
catch (Exception e)
{
throw new Exception($"Column name {fields.Count} is not valid UTF-8", e);
}
if (columnString.Contains(':'))
{
throw new Exception($"Header {fields.Count} contain ':', which is not allowed for column names");
}
columnNames[j] = columnString;
}
fields.Clear();
}
else 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($"Found unescaped '#' at line {line}, column {i - currentLineStart}");
}
else
{
fieldBytes.Add(inputBuffer[i]);
}
}
fields.Add(fieldBytes.ToArray());
if (numFields == 0)
{
throw new Exception("Found 0 fields on last line. Possibly because of extra \\n after last record");
}
if (numFields != fields.Count)
{
throw new Exception($"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($"Line {line}, column {j} is not valid UTF-8", e);
}
}
records.Add(fieldStrings);
fields.Clear();
}
return (columnNames, records.ToArray());
}
public static (string[] columns, string[][] data) ParseSimpleTsvParallel(byte[] inputBuffer)
{ {
string[] columnNames = null; string[] columnNames = null;
var headers = new List<byte[]>(); var headers = new List<byte[]>();
@ -1423,10 +930,10 @@ public class SaneTsv
} }
} }
return (columnNames, ParseSimpleTsvParallel(inputBuffer, columnNames.Length, startOfData, inputBuffer.Length)); return (columnNames, ParseSimpleTsv(inputBuffer, columnNames.Length, startOfData, inputBuffer.Length));
} }
public static string[][] ParseSimpleTsvParallel(byte[] inputBuffer, int numFields, int startIndex, int endIndex) public static string[][] ParseSimpleTsv(byte[] inputBuffer, int numFields, int startIndex, int endIndex)
{ {
var fieldBytes = new List<byte>(); var fieldBytes = new List<byte>();
var fields = new List<byte[]>(); var fields = new List<byte[]>();
@ -1666,282 +1173,7 @@ public class SaneTsv
return SerializeTsv<T>(data, FormatType.COMMENTED_TSV); return SerializeTsv<T>(data, FormatType.COMMENTED_TSV);
} }
protected static byte[] SerializeTsv<T>(IList<T> data, FormatType tsvFormat) where T : TsvRecord protected static byte[] SerializeTsv<T>(IList<T> data, FormatType tsvFormat)
{
var bytes = new List<byte>();
var headerTypes = new List<Type>();
var headerNames = new List<string>();
var headerPropertyInfos = new List<PropertyInfo>();
int columnCount = 0;
foreach (PropertyInfo property in typeof(T).GetProperties())
{
TsvColumnAttribute attribute = (TsvColumnAttribute)Attribute.GetCustomAttribute(property, typeof(TsvColumnAttribute));
if (attribute == null)
{
continue;
}
string headerName = attribute.ColumnName ?? property.Name;
headerNames.Add(headerName);
Type headerType = attribute.ColumnType ?? GetColumnFromType(property.PropertyType);
if (tsvFormat == FormatType.SIMPLE_TSV && headerType != typeof(StringType))
{
throw new Exception($"Serializing Simple TSV requires all columns be of type string, but column '{headerName}' has type '{headerType}'");
}
headerTypes.Add(headerType);
headerPropertyInfos.Add(property);
// TODO: Check that the property type and given column type are compatible
columnCount++;
}
// Serialize header
for (int i = 0; i < headerNames.Count; i++)
{
for (int j = i + 1; j < headerNames.Count; j++)
{
if (headerNames[i] == headerNames[j])
{
throw new Exception("Column names in header must be unique");
}
}
byte[] nameEncoded = Encoding.UTF8.GetBytes(headerNames[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(headerTypes[i])));
}
catch (Exception e)
{
throw new Exception($"Invalid header type for column {i}", e);
}
}
if (i == headerNames.Count - 1)
{
bytes.Add((byte)'\n');
}
else
{
bytes.Add((byte)'\t');
}
}
// Serialize data
for (int i = 0; i < data.Count; i++)
{
for (int j = 0; j < columnCount; j++)
{
object datum = headerPropertyInfos[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 (headerTypes[j] == typeof(StringType))
{
fieldEncoded = Encoding.UTF8.GetBytes((string)datum);
}
else if (headerTypes[j] == typeof(BooleanType))
{
bytes.AddRange((bool)datum ? TrueEncoded : FalseEncoded);
skipEscaping = true;
}
else if (headerTypes[j] == 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 (headerTypes[j] == 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 (headerTypes[j] == typeof(Float64Type))
{
if (datum is double d)
{
if (double.IsNegativeInfinity(d))
{
bytes.AddRange(Encoding.UTF8.GetBytes("-inf"));
}
else if (double.IsPositiveInfinity(d))
{
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((d).ToString("G17")));
}
}
else
{
throw new InvalidCastException();
}
skipEscaping = true;
}
else if (headerTypes[j] == typeof(Float64LEType))
{
if (LittleEndian)
{
fieldEncoded = BitConverter.GetBytes((double)datum);
}
else
{
byte[] doubleBytes = BitConverter.GetBytes((double)datum);
fieldEncoded = new byte[sizeof(double)];
for (int k = 0; k < sizeof(double); k++)
{
fieldEncoded[k] = doubleBytes[sizeof(double) - 1 - k];
}
}
}
else if (headerTypes[j] == typeof(UInt32Type))
{
bytes.AddRange(Encoding.UTF8.GetBytes(((UInt32)datum).ToString()));
skipEscaping = true;
}
else if (headerTypes[j] == typeof(UInt64Type))
{
bytes.AddRange(Encoding.UTF8.GetBytes(((UInt64)datum).ToString()));
skipEscaping = true;
}
else if (headerTypes[j] == typeof(Int32Type))
{
bytes.AddRange(Encoding.UTF8.GetBytes(((Int32)datum).ToString()));
skipEscaping = true;
}
else if (headerTypes[j] == typeof(Int64Type))
{
bytes.AddRange(Encoding.UTF8.GetBytes(((Int64)datum).ToString()));
skipEscaping = true;
}
else if (headerTypes[j] == typeof(BinaryType))
{
fieldEncoded = (byte[])datum;
}
else
{
throw new Exception($"Unexpected column type {headerTypes[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 < columnCount - 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(headerTypes[j])}", e);
}
}
}
return bytes.ToArray();
}
protected static byte[] SerializeTsvParallel<T>(IList<T> data, FormatType tsvFormat)
{ {
var bytes = new List<byte>(); var bytes = new List<byte>();
@ -2037,12 +1269,12 @@ public class SaneTsv
} }
// Serialize data // Serialize data
SerializeTsvParallel<T>(data, bytes, headerPropertyInfos.ToArray(), headerTypes.ToArray(), tsvFormat, 0, data.Count); SerializeTsv<T>(data, bytes, headerPropertyInfos.ToArray(), headerTypes.ToArray(), tsvFormat, 0, data.Count);
return bytes.ToArray(); return bytes.ToArray();
} }
protected static void SerializeTsvParallel<T>(IList<T> data, List<byte> bytes, PropertyInfo[] headerPropertyInfos, Type[] headerTypes, FormatType tsvFormat, int startIndex, int endIndex) protected static void SerializeTsv<T>(IList<T> data, List<byte> bytes, PropertyInfo[] headerPropertyInfos, Type[] headerTypes, FormatType tsvFormat, int startIndex, int endIndex)
{ {
// Serialize data // Serialize data
for (int i = 0; i < data.Count; i++) for (int i = 0; i < data.Count; i++)

216
SaneTsv/SaneTsvTest/Program.cs
View File

@ -348,222 +348,6 @@ internal class Program : SaneTsv
Console.WriteLine($"Unspecced parse time: {unspeccedParseTime}"); Console.WriteLine($"Unspecced parse time: {unspeccedParseTime}");
} }
{
string testName = "Check parallel Simple TSV serialization";
int N = 100000;
var records = new StringTestRecord[N];
var rand = new Random(1);
for (int i = 0; i < N; i++)
{
records[i] = new StringTestRecord()
{
Column1 = rand.Next().ToString(),
column2 = rand.Next().ToString(),
Column3 = rand.Next().ToString(),
};
}
string[][] recordStrings = records.Select(record => new string[] { record.Column1, record.column2, record.Column3 }).ToArray();
DateTime lastTime = DateTime.Now;
byte[] serialized1 = SaneTsv.SerializeSimpleTsv(new string[] { "column1", "column2", "columnthree\nyep" }, recordStrings);
TimeSpan unparallelTime = DateTime.Now - lastTime;
lastTime = DateTime.Now;
byte[] serialized2 = SaneTsv.SerializeSimpleTsvParallel(new string[] { "column1", "column2", "columnthree\nyep" }, recordStrings);
TimeSpan parallelTime = DateTime.Now - lastTime;
Console.WriteLine($"Unparallel serialization time: {unparallelTime}");
Console.WriteLine($"Parallel serialization time: {parallelTime}");
bool matching = true;
for (int i = 0; i < Math.Min(serialized1.Length, serialized2.Length); i++)
{
if (serialized1[i] != serialized2[i])
{
matching = false;
break;
}
}
if (matching)
{
Console.WriteLine($"Passed {testName}");
}
else
{
Console.WriteLine($"Failed {testName}");
}
}
{
string testName = "Check Simple TSV parallel parsing";
int N = 100000;
var records = new StringTestRecord[N];
var rand = new Random(1);
for (int i = 0; i < N; i++)
{
records[i] = new StringTestRecord()
{
Column1 = rand.Next().ToString(),
column2 = rand.Next().ToString(),
Column3 = rand.Next().ToString(),
};
}
byte[] serialized = SaneTsv.SerializeSimpleTsv<StringTestRecord>(records);
DateTime lastTime = DateTime.Now;
(string[] headers2, string[][] data2) = SaneTsv.ParseSimpleTsv(serialized);
TimeSpan unparallelTime = DateTime.Now - lastTime;
lastTime = DateTime.Now;
(string[] headers, string[][] data) = SaneTsv.ParseSimpleTsvParallel(serialized);
TimeSpan parallelTime = DateTime.Now - lastTime;
Console.WriteLine($"Unparallel parse time: {unparallelTime}");
Console.WriteLine($"Parallel parse time: {parallelTime}");
bool matching = true;
for (int j = 0; j < Math.Min(headers2.Length, headers.Length); j++)
{
if (headers[j] != headers2[j])
{
matching = false;
break;
}
}
for (int i = 0; i < Math.Min(data.Length, data2.Length) && matching; i++)
{
for (int j = 0; j < data[0].Length; j++)
{
if (data[i][j] != data2[i][j])
{
matching = false;
break;
}
}
}
if (matching)
{
Console.WriteLine($"Passed {testName}");
}
else
{
Console.WriteLine($"Failed {testName}");
}
}
{
string testName = "Check parallel serialization";
int N = 1000;
var records = new BoolTestRecord[N];
var rand = new Random(1);
for (int i = 0; i < N; i++)
{
byte[] bytes = new byte[rand.Next(50)];
rand.NextBytes(bytes);
records[i] = new BoolTestRecord()
{
Column1 = rand.NextDouble() > 0.5,
column2 = bytes,
Column3 = rand.Next().ToString(),
};
}
DateTime lastTime = DateTime.Now;
byte[] serialized1 = SaneTsv.SerializeTsv<BoolTestRecord>(records, FormatType.COMMENTED_TSV);
TimeSpan unparallelTime = DateTime.Now - lastTime;
lastTime = DateTime.Now;
byte[] serialized2 = SaneTsv.SerializeTsvParallel<BoolTestRecord>(records, FormatType.COMMENTED_TSV);
TimeSpan parallelTime = DateTime.Now - lastTime;
Console.WriteLine($"Unparallel serialization time: {unparallelTime}");
Console.WriteLine($"Parallel serialization time: {parallelTime}");
bool matching = true;
for (int i = 0; i < Math.Min(serialized1.Length, serialized2.Length); i++)
{
if (serialized1[i] != serialized2[i])
{
matching = false;
break;
}
}
if (matching)
{
Console.WriteLine($"Passed {testName}");
}
else
{
Console.WriteLine($"Failed {testName}");
}
}
{
string testName = "Check parallel parsing";
int N = 1000;
var records = new BoolTestRecord[N];
var rand = new Random(1);
for (int i = 0; i < N; i++)
{
byte[] bytes = new byte[rand.Next(50)];
rand.NextBytes(bytes);
records[i] = new BoolTestRecord()
{
Column1 = rand.NextDouble() > 0.5,
column2 = bytes,
Column3 = rand.Next().ToString(),
};
}
byte[] serialized2 = SaneTsv.SerializeTsvParallel<BoolTestRecord>(records, FormatType.COMMENTED_TSV);
DateTime lastTime = DateTime.Now;
CommentedTsv<BoolTestRecord> parsed = (CommentedTsv<BoolTestRecord>)SaneTsv.Parse<BoolTestRecord>(serialized2, FormatType.COMMENTED_TSV);
TimeSpan unparallelTime = DateTime.Now - lastTime;
lastTime = DateTime.Now;
CommentedTsv<BoolTestRecord> parsed2 = (CommentedTsv<BoolTestRecord>)SaneTsv.ParseParallel<BoolTestRecord>(serialized2, FormatType.COMMENTED_TSV);
TimeSpan parallelTime = DateTime.Now - lastTime;
Console.WriteLine($"Unparallel parsing time: {unparallelTime}");
Console.WriteLine($"Parallel parsing time: {parallelTime}");
bool matching = parsed.FileComment == parsed2.FileComment;
matching &= parsed.Records.Count == parsed2.Records.Count;
for (int i = 0; matching && i < parsed.Records.Count; i++)
{
matching &= parsed.Records[i].Comment == parsed2.Records[i].Comment;
matching &= parsed.Records[i].Column1 == parsed2.Records[i].Column1;
matching &= parsed.Records[i].column2.Length == parsed2.Records[i].column2.Length;
for (int j = 0; matching && j < parsed.Records[i].column2.Length; j++)
{
matching &= parsed.Records[i].column2[j] == parsed2.Records[i].column2[j];
}
}
if (matching)
{
Console.WriteLine($"Passed {testName}");
}
else
{
Console.WriteLine($"Failed {testName}");
}
}
Console.WriteLine("Done with tests"); Console.WriteLine("Done with tests");
} }
} }