Reference

Version 2.4.1. Anything not listed here is not part of the public API.

Layout and Field

`Name`	Field name. Must be unique within a layout; used in lookups and errors only.
`Width`	Field width in bytes. Must be greater than zero.
`Type`	One of the field types below.
`Scale`	Implied decimal places. `Decimal` only.
`Format`	Reference layout string. `Timestamp` and `Date` only.
`Pad`	`SpaceRight` (default), `SpaceLeft`, `ZeroLeft`.
`Optional`	An all-pad field decodes to the zero value instead of an error.

Layout.RecordLength() returns the sum of the widths. Layout.Validate() reports a duplicate name, a zero width, a scale on a non-decimal field, or a missing format on a temporal field. Both readers and writers call Validate on construction, so an invalid layout fails before any I/O happens.

Field types

`Text`	Bytes decoded with the reader's encoding, padding trimmed according to `Pad`.
`Integer`	Optionally signed decimal digits. An overpunched sign in the final byte is accepted.
`Decimal`	Digits with an implied decimal point at `Scale`. Exact; never a float.
`Packed`	Packed decimal, two digits per byte, sign in the low nibble.
`Date`	Calendar date, no time zone.
`Timestamp`	Date and time. UTC unless the format carries an offset.
`Raw`	Bytes, untouched. Padding is not trimmed and no encoding is applied.

Reader

`NewReader(io.Reader, Layout) *Reader`	Constructs a reader. Panics only if the layout is nil.
`Next() bool`	Decodes the next record. False at EOF and on the first error.
`Record() Record`	The record decoded by the last `Next`. Invalid after the following call.
`Err() error`	The first error encountered, or nil at a clean EOF.
`Offset() int64`	Byte offset of the start of the current record.
`Encoding`	`ASCII` (default), `Latin1`, `UTF8`.
`Terminator`	`None` (default), `LF`, `CRLF`, `Auto`.
`BufferSize`	Read buffer in bytes, default 262144, rounded up to whole records.
`OnError`	`func(*PositionError) Action`. Return `Skip` or `Stop`.

Record

`Text(name) (string, error)`	Decoded text. Error if the field is not `Text`.
`Int(name) (int64, error)`	Integer value. Error on any non-digit byte in a required field.
`Decimal(name) (Decimal, error)`	Exact decimal with the declared scale.
`Time(name) (time.Time, error)`	Parsed with the field's format.
`Bytes(name) ([]byte, error)`	The raw field bytes. Aliases the reader's buffer.
`IsBlank(name) bool`	True if the field is entirely padding.
`Clone() Record`	An independent copy, safe to retain.

Writer

`NewWriter(io.Writer, Layout) *Writer`	Constructs a buffered writer.
`Write(Values) error`	Encodes one record. A missing required field is an error.
`Flush() error`	Writes buffered records to the underlying writer.
`Close() error`	Flushes and marks the writer used. Must be checked.
`Terminator`	Matches the reader's; default `None`.

Errors

`ErrShortRecord`	The stream ended part-way through a record.
`ErrFieldSyntax`	The bytes of a field do not match its declared type.
`ErrFieldOverflow`	A value written does not fit its declared width.
`ErrLayout`	The layout is self-inconsistent; returned before any I/O.
`ErrEncoding`	A byte sequence is not valid in the declared encoding.
`PositionError`	Wraps any of the above with `Offset`, `Record`, `Field`.

All sentinel errors are comparable with errors.Is through the wrapping PositionError. New sentinels may be added in a minor release; existing ones will not be removed before 3.0.