Monday, September 21, 2015

Emacspeak: An Overview Of Voice-Lock Over The Years

Emacspeak: An Overview Of Voice-Lock Over The Years

1 Executive Summary

Audio formatted output is provided via voice-lock in Emacspeak.
The feature was originally implemented in 1994 and significantly
overhauled in 2002 (with no change in functionality). Now, in 2015,
the implementation has been reworked to be once again simpler — but
with the same functionality.


2 1994: Original Implementation

Font-lock was a new feature in Emacs 19 — it was introduced right
around the time the Emacspeak project was started.
At the time, there were multiple font-lock packages in
Emacs. Emacspeak implemented voice-lock via a stand-alone module
that attached property personality to buffer contents. The
dtk-speak Text-To-Speech module treated strings annotated with
property personality as an aural display list when producing
spoken output.


In the winter of 1995, I implemented Aural CSS in emacspeak, and
this resulted in module acss.el — that module provides a clean
abstraction for defining values that can be assigned to property
personality. Next, the rest of emacspeak was overhauled to
express all voice-lock features via the ACSS abstraction.


3 2002 Overhaul

With feature font-lock now mature, Emacs itself converged on a
single font-lock module with jit-lock.el providing just-in-time font
locking. This meant that Emacspeak's voice-lock feature could now be
more easily implemented by attaching appropriate advice to lisp
functions put-text-property and friends — this led to the deletion
of module voice-lock.el and the creation of module
emacspeak-personality.el.


That module provided multiple options for how faces were mapped to
personalities:


  • Cumulative where property personality was either appended or
    prepended to the list of personalities at any given position.
  • Simple: where the new personality directly replaced any previously
    applied personality property.
  • None: Where faces were not mapped to personalities.

4 2015 Overhaul

And now, in 2015, the code in module dtk-speak.el that implements
audio-formatting has been overhauled to directly map faces to
personalities. This implementation can be enabled by setting the
option emacspeak-personality-voiceify-faces provided in module
emacspeak-personality.el to All. This means that the new
implementation can co-exist (at least for a while) with the earlier
advice-based implementation from module emacspeak-personality.el.


once deemed to be stable, the new implementation will become the
default — by changing the default value of
emacspeak-personality-voiceify-faces to none.

5 Summary

From an end-user's perspective, nothing has changed in 21 years. From
the perspective of implementing voice-lock for Emacs modules, nothing
has changed — all that an Emacspeak extension for a specific package
needs to do to enable voice-lock is to set up a face->personality
mapping as before via function voice-setup-add-map. Note that
property personality will continue to have higher precedence than
property face with respect to the audio formatting that gets applied
i.e. Emacspeak first consults value of property personality, and if
that is not set, but there is a face property set, then that
face->personality mapping (if defined) is used as the aural style.

Date: <2015-08-18 Tue>

Author: T.V Raman

Created: 2015-08-21 Fri 10:18

Validate