branch: master
commit bfdb6403ea77f285f381f76d179fd5d105e39b46
Author: Ian Dunn <du...@gnu.org>
Commit: Ian Dunn <du...@gnu.org>
Updated pabbrev comparison
* paced.org (pabbrev): Clarify difficulty with working with files that
aren't
always open.
---
paced.org | 27 +++++++++++++--------------
1 file changed, 13 insertions(+), 14 deletions(-)
diff --git a/paced.org b/paced.org
index 78bef8b..3b5abb4 100644
--- a/paced.org
+++ b/paced.org
@@ -56,23 +56,22 @@ completion can be customized.
There are a few Emacs packages that have similar goals to paced, and provided
some of the inspiration and motivation behind it.
*** pabbrev
-The [[https://github.com/phillord/pabbrev][pabbrev]] package by Phillip Lord
scans text of the current buffer while Emacs
-is idle and presents the user with the most common completions.
+The [[https://github.com/phillord/pabbrev][pabbrev]] package by Phillip Lord
automatically scans text of the current
+buffer while Emacs is idle and presents the user with the most common
+completions.
One of the major downsides to pabbrev is that the data it collects doesn't
persist between Emacs sessions. For a few files that are always open, such as
-org agenda files, pabbrev works great. For files that aren't always open, like
-prose or source files, you've got to retrain pabbrev every time you restart
-Emacs.
-
-The benefit of pabbrev dies down if the suggested word isn't the one you need.
-Then you've still got to search through a list of suggestions, which takes away
-from typing.
-
-That's not to say that pabbrev is all bad. It keeps up-to-date usage and
prefix
-hashes of all buffers of the same mode, and scanning, or "scavenging", blends
-seamlessly into the background. Completion is just a hash table lookup, so it
-can handle completion in microseconds.
+org agenda files, pabbrev works great. If you want to train it from a few
files
+that aren't always open, you'll have to open each file and retrain pabbrev from
+that file. And you'll have to do this every time you restart Emacs.
+
+It keeps up-to-date usage and prefix hashes of all buffers of the same mode,
and
+scanning, or "scavenging", blends seamlessly into the background. Completion
is
+just a hash table lookup, so it can handle completion in microseconds. There's
+also no setup required; it will start working right away. The downside to this
+is that dictionaries aren't flexible; each dictionary corresponds to a major
+mode, and there's no way to change that.
*** predictive
The [[https://www.dr-qubit.org/predictive.html][predictive]] package by Toby
Cubitt scans text of the current buffer on user
command. The usage data is stored in a dictionary, which can then be saved to
a
