Fiff Library (`FIFFLIB`)

The Fiff library implements the FIFF (Functional Image File Format) used throughout the MNE ecosystem for MEG/EEG data. It provides file I/O, in-memory data containers, channel and sensor metadata, digitisation points, coordinate transforms, SSP projections, and CTF compensation — everything required to read, manipulate, and write neurophysiological recordings. All classes reside in the FIFFLIB namespace and depend only on Qt and Eigen.

Architecture

Class Inventory

File I/O

Class	Description	MNE-Python	MNE-C
`FiffStream`	Low-level tag-based FIFF file reader/writer — open, read/write individual tags, navigate directory tree	`mne._fiff.open.fiff_open()`	`fiffFile`
`FiffIO`	High-level convenience reader/writer for complete measurement files	`mne.io.read_raw_fif()`	`mne_read_*()` family
`FiffTag`	Single data tag: kind, type, size, position, and typed payload	`mne._fiff.tag.Tag`	`fiffTag`
`FiffDirEntry`	Directory entry describing a tag's location within the file	`mne._fiff.tree`	`fiffDirEntry`
`FiffDirNode`	Tree node representing a block in the FIFF directory hierarchy (e.g., `FIFFB_MEAS`, `FIFFB_MNE_ENV`)	`mne._fiff.tree.dir_tree_find()`	`fiffDirTree`
`FiffId`	Universally unique file/block identifier (UUID + timestamp)	`mne._fiff.tag`	`fiffId`

Data Containers

Class	Description	MNE-Python	MNE-C
`FiffRawData`	Raw continuous MEG/EEG data with lazy reading, ring-buffer support, projection, and compensation	`mne.io.Raw`	`mneRawData`
`FiffEvoked`	Averaged/evoked response data (channels × time) with stimulus metadata	`mne.Evoked`	`mneEvoked`
`FiffEvokedSet`	Collection of multiple evoked data sets read from a single file	`mne.read_evokeds()`	—
`FiffCov`	Noise or data covariance matrix (dense or diagonal) with eigendecomposition and channel names	`mne.Covariance`	`mneCov`
`FiffEvents`	Event array (sample × 3 matrix: sample, value-before, value-after)	`mne.read_events()`	—
`FiffNamedMatrix`	Matrix with labelled rows and columns, shared-data semantics	Internal	`mneNamedMatrix`
`FiffSparseMatrix`	Sparse matrix in Compressed Column/Row Storage	`scipy.sparse` equivalent	`mneSparseMatrix`

Channel & Sensor Information

Class	Description	MNE-Python	MNE-C
`FiffInfo`	Complete measurement metadata: channels, sampling rate, SSP projections, CTF data, coordinate transforms, subject info, date	`mne.Info`	`mneMeasData`
`FiffInfoBase`	Lightweight base class for measurement info (channel list, bads, sampling rate)	`mne.Info` (base interface)	—
`FiffChInfo`	Per-channel metadata: name, type (MEG/EEG/STIM/…), unit, calibration, location	`mne.Info['chs'][i]`	`fiffChInfo`
`FiffChPos`	Sensor position and orientation: integration-point location + two ex/ey vectors	`mne.Info['chs'][i]['loc']`	`fiffChPos`

Digitisation & Spatial Data

Class	Description	MNE-Python	MNE-C
`FiffDigPoint`	Single digitised point: kind (cardinal/HPI/EEG/extra), ident, 3D position	`mne._fiff._digitization.DigPoint`	`fiffDigPoint`
`FiffDigPointSet`	Ordered collection of digitiser points with coordinate transform support	`mne.Info['dig']`	—
`FiffDigitizerData`	Digitiser point set with associated head-to-device transformation	`mne.Info['dig']` + transforms	—
`FiffCoordTrans`	Rigid-body coordinate transformation (4 × 4 homogeneous matrix) between two frames	`mne.transforms.Transform`	`fiffCoordTrans`
`FiffCoordTransSet`	Chain of transforms (e.g., head → MRI → MNI Talairach)	`mne.transforms` collection	—

Calibration & Correction

Class	Description	MNE-Python	MNE-C
`FiffProj`	Signal Space Projection (SSP) operator — one projection vector with kind and description	`mne.Projection`	`mneProj`
`FiffCtfComp`	CTF software gradient compensation data: compensation matrix and calibration	`mne.Info['comps']`	`mneCtfComp`

Key utility methods on FiffInfo:

make_compensator() — create a compensation matrix for a given grade
get_current_comp() — query active CTF compensation setting
make_projector() — build an SSP projector from active projections
pick_types() / pick_channels() — channel selection by type or name

Constants & Types

Header	Description	MNE-Python	MNE-C
`fiff_constants.h`	All FIFF block, tag, and value identifiers (`FIFFB_`, `FIFF_`, `FIFFV_*`)	`mne._fiff.constants.FIFF`	`mne_fiff.h`
`fiff_types.h`	Primitive FIFF types: `fiff_int_t`, `fiff_float_t`, `fiff_byte_t`, DAU pack types	—	`mne_types.h`
`fiff_explain.h`	Human-readable string maps for FIFF constants (block type → name, tag kind → name)	—	—

Usage Example

#include <fiff/fiff.h>

using namespace FIFFLIB;

// Open a raw FIFF file
QFile file("sample_audvis_raw.fif");
FiffStream::SPtr stream(new FiffStream(&file));
FiffRawData raw(stream);

// Query measurement info
qDebug() << "Sample rate:" << raw.info.sfreq;
qDebug() << "Channels:"   << raw.info.nchan;
qDebug() << "Bad channels:" << raw.info.bads;

// Read a time window [0 s, 10 s]
Eigen::MatrixXd data;
Eigen::RowVectorXd times;
raw.read_raw_segment(data, times, 0, raw.info.sfreq * 10);

// Apply SSP projectors
FiffProj::activate_projs(raw.info.projs);
Eigen::MatrixXd proj;
int nproj;
FiffProj::make_projector(raw.info.projs, raw.info.ch_names, proj, nproj);
data = proj * data;

// Read evoked data
QList<FiffEvoked> evokeds = FiffEvoked::read("sample_audvis-ave.fif");

// Read noise covariance
FiffCov cov = FiffCov::read("sample_audvis-cov.fif");

Doxygen Reference

For method signatures, inheritance diagrams, and source-level documentation see the auto-generated FIFFLIB namespace in the Doxygen API reference.

Architecture​

Class Inventory​

File I/O​

Data Containers​

Channel & Sensor Information​

Digitisation & Spatial Data​

Calibration & Correction​

Constants & Types​

Usage Example​

Doxygen Reference​

See Also​