Install PocketCsvReader.KeyValue to read records whose fields carry their own labels. Use LogfmtReaderBuilder for space-separated key=value records or LtsvReaderBuilder for tab-separated label:value records.
dotnet add package PocketCsvReader.KeyValue
Example: read application logs with logfmt
This complete example reads two application log entries. The second entry deliberately changes the field order, so the reader resolves labels on each row.
using System;
using System.IO;
using System.Text;
using PocketCsvReader.KeyValue.Configuration;
var content = "level=info message=\"Order accepted\" order_id=1001\n"
+ "order_id=1002 message=\"Payment pending\" level=warn\n";
var logs = new LogfmtReaderBuilder().Build();
using var stream = new MemoryStream(Encoding.UTF8.GetBytes(content));
using var reader = logs.ToDataReader(stream);
while (reader.Read())
{
var level = reader.GetString(reader.GetOrdinal("level"));
var orderId = reader.GetInt32(reader.GetOrdinal("order_id"));
var message = reader.GetString(reader.GetOrdinal("message"));
Console.WriteLine($"{level}: order {orderId}, {message}");
}
Output:
info: order 1001, Order accepted
warn: order 1002, Payment pending
To read a file, replace logs.ToDataReader(stream) with logs.ToDataReader("application.log").
Logfmt values can be unquoted or double-quoted. Quoted values support escapes such as \", \\, and \n. GetString returns the decoded value; GetRawString retains the original field value, including its quotes and escapes. Bare keys such as active and explicit empty values such as message= return empty strings.
Example: read LTSV request logs
LTSV separates fields with a tab and splits each field at its first colon. A value can contain further colons, as in a URL or timestamp. This example includes the same labels in different orders:
using System;
using System.IO;
using System.Text;
using PocketCsvReader.KeyValue.Configuration;
var content = "host:example.org\tstatus:200\tpath:/orders/1001\n"
+ "path:/orders/1002\thost:example.org\tstatus:404\n";
var requests = new LtsvReaderBuilder().Build();
using var stream = new MemoryStream(Encoding.UTF8.GetBytes(content));
using var reader = requests.ToDataReader(stream);
while (reader.Read())
{
var status = reader.GetInt32(reader.GetOrdinal("status"));
var path = reader.GetString(reader.GetOrdinal("path"));
Console.WriteLine($"{status}: {path}");
}
Output:
200: /orders/1001
404: /orders/1002
Labels and changing record shapes
Labels come from each record, rather than a header row. Call Read() before looking up a field. FieldCount, field order, and labels may change between records; resolve ordinals on the current row instead of caching them across rows.
Use TryGetOrdinal("label", out var ordinal) to check for optional fields before accessing them. Duplicate labels are preserved, and GetOrdinal returns the first matching occurrence. Access subsequent occurrences by index. A missing label passed to GetOrdinal throws ArgumentOutOfRangeException.
Both formats accept LF, CRLF, or CR line endings and a final record without a newline. An empty line is exposed as a record with zero fields. Invalid syntax throws InvalidDataException.
Configuration and output models
Both builders support the shared WithSchema, WithResource, WithParsers, and WithParserOptimizations settings. Use a named schema when field types should follow labels as their positions change. See Schema configuration and Resource configuration for details.
The readers expose ToDataReader, ToDbDataReader, ToDataTable, and ToArrayString, with filename and stream overloads. ToDataReader is useful for variable record shapes; tabular output requires a consistent set of columns. They also provide the shared parsed-row materializers such as ToDictionaries and ToObjectArrays.
Input is processed in one forward-only pass, including non-seekable streams. Read-ahead and row-count pre-scans are disabled for these formats. Keep the reader open while accessing its current row, and dispose it when finished. Materialized strings and output objects remain valid after the next read.
For null versus empty values see value handling; for reused buffers and metadata see key-value buffering.
