Track fixations across edits for eyetracking experiments
gazel provides fixation algorithms for aggregating gazes, as well as tools to help you deal with edits that occur during eyetracking sessions.
gazel requires python 3.8.5
You can install gazel using
pip install gazel
gazel comes with a command line interface as well as a python library.
To use either version, you need 3 things:
- itrace core file output
- gaze file output
- original source file(s)
To get started:
# import gazel from gazel import fixation_filter, Tracker # run fixation filter to get fixations fixations = fixation_filter(gazes, changelog, sources, **opts) # Now track edits using Tracker tracker = Tracker(fixations, changelog, sources, language) # Note that you need to pass in the source code language, # to help `gazel` determine which parser to use.
The main goal of
Tracker is to track fixations and source code tokens across edits. Once you create a
Tracker you can query it to get snapshots etc.
Tracker maintains a list of snapshots corresponding to the original file and subsequent edits. The original version is stored at index
0 and the first edit is stored at index
# get original tracker.get_snapshot(0) # get first edit tracker.get_snapshot(1)
Each snapshot is represented by a
Snapshot is defined as:
class Snapshot: id: int source: Source tokens: Tuple[Token, ...] changes: Tuple[TokenChange, ...] = () time: float = 0.0
Snapshot.time represents the time at which a snapshot was created. It corresponds to the timestamp in the changelog that was used to create this
Snapshot.tokens represents the parsed source code tokens.
Snapshot.changes represents all the token changes that happened to this
Snapshot since the last version. For the
Snapshot representing the original source,
Snapshot.changes is empty.
Snapshots.source is a
Source object, containing the raw text of the source code, as well as mappings from text indices to line/column numbers and vice-versa. It is defined as follows:
@dataclass(frozen=True) class Source: text: str mapping: PositionMapping language: str
You can retreive gazes for a given time window as follows:
tracker.get_gazes() # all gazes tracker.get_gazes(start, end) # returns a dataframe that is filtered
The gaze dataframe is simply a
pandas.DataFrame containing the original gazes, with some additional columns:
syntax_node- The syntax node associated with the gaze.
Noneif the gaze doesn't fall on a token.
syntax_node_id- A stable id for the token associated with this gaze.
Noneif the gaze doesn't fall on a token.
syntax_node_id is a unique id that is assigned to each token in the source code across different snapshots. For a given token, this id is unique across time and space. Thus, you can use this id to determine how
tracker.diff(0, 2) # gives you the diff between version 0 & version 1
tracker.diff_time(2300, 2400) # gives you the diff between time unit 2300 & time unit 2400
SnapshotDiff is defined as follows:
class SnapshotDiff(NamedTuple): old: Snapshot new: Snapshot token_changes: List[TokenChange] gaze_changes: List[GazeChange] gazes: pd.DataFrame
It gives you a list of all token changes and gaze changes.
Token changes can be of 3 types:
Gaze changes can be of 2 types:
deleted means that the token to which the gaze was mapped to has been removed from the source.)
tracker.diff(start, end) will include all the gazes from the start of the experiment (the time at which the first gaze is recorded), up until the
end snapshot time. If you want to only include gazes within the timespan of the
end snapshots, you can pass an optional parameter:
diff = tracker.diff(0, 3, window_only=True)
gazel also provides a pretty-printer to help you print diffs for inspecting the data. It supports print gaze changes, token changes and
from gazel import pprint diff = tracker.diff(2, 3) pprint(diff) pprint(diff.token_changes) pprint(diff.gaze_changes)
gazel provides a module
transforms to help you manipulate
gazel structures including
from gazel import transforms as T diff = tracker.diff(2, 5) # TODO
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.