elfeed-summary
elfeed-summary
The package provides a tree-based feed summary interface for elfeed. The tree can include individual feeds, searches, and groups. It mainly serves as an easier “jumping point” for elfeed, so to make querying a subset of the elfeed database one action away.
Inspired by newsboat.
Installation
The package is available on MELPA, so install it however you normally install packages. My preferred way is use-package
with straight
:
(use-package elfeed-summary
:straight t)
Of course, you have to have elfeed configured.
Usage
Running M-x elfeed-summary
opens up the summary buffer, as shown on the screenshot.
The tree consists of:
- feeds;
- searches;
- groups, that can include other groups, feeds, and searches.
Groups can also be generated automatically.
Available keybindings in the summary mode:
Keybinding | Command | Description |
---|---|---|
RET |
elfeed-summary--action |
Open thing under the cursor (a feed, search, or a group). If there is at least one unread item, it will show only unread items. |
M-RET |
elfeed-summary--action-show-read |
Open thing under the cursor, but always include read items |
q |
… | Quit the summary buffer |
r |
elfeed-summary--refresh |
Refresh the summary buffer |
R |
elfeed-summary-update |
Run update for elfeed feeds |
u |
elfeed-summary-toggle-only-unread |
Toggle showing only unread entries |
U |
elfeed-summary--action-mark-read |
Mark everything in the entry under the cursor as read |
The standard keybindings from magit-section are also available, for instance TAB
toggles the visibility of the current group. evil-mode is also supported.
Configuration
Tree configuration
The structure of the tree is determined by the elfeed-summary-settings
variable.
This is a list of these possible items:
- Group
(group . <group-params>)
Groups are used to group elements under collapsible sections. - Query
(query . <query-params>)
Query extracts a subset of elfeed feeds based on the given criteria. Each found feed will be represented as a line. - Search
(search . <search-params>)
Elfeed search, as defined byelfeed-search-set-filter
. - Tags tree
(auto-tags . <auto-tags-params>)
A tree generated automatically from the available tags. - Tag groups
(tag-groups . <tag-group-params>)
Insert one tag as one group. - a few special forms
<group-params>
is an alist with the following keys:
:title
(mandatory):elements
(mandatory) - elements of the group. The structure is the same as in the root definition.:face
- group face. The default face iselfeed-summary-group-face
.:hide
- if non-nil, the group is collapsed by default.
<query-params>
can be:
- A symbol of a tag. A feed will be matched if it has that tag.
:all
. Will match anything.(title . "string")
or(title . <form>)
Match feed title withstring-match-p
. <form> makes sense if you want to pass something likerx
.(author . "string")
or(author . <form>)
(url . "string")
or(url . <form>)
(and <q-1> <q-2> ... <q-n>)
Match if all the conditions 1, 2, …, n match.(or <q-1> <q-2> ... <q-n>)
or(<q-1> <q-2> ... <q-n>)
Match if any of the conditions 1, 2, …, n match.(not <query>)
Feed tags for the query are determined by the elfeed-feeds
variable.
Query examples:
(emacs lisp)
Return all feeds that have either “emacs” or “lisp” tags.(and emacs lisp)
Return all feeds that have both “emacs” and “lisp” tags.(and (title . "Emacs") (not planets))
Return all feeds that have “Emacs” in their title and don’t have the “planets” tag.
<search-params>
is an alist with the following keys:
:filter
(mandatory) filter string, as defined byelfeed-search-set-filter
:title
(mandatory) title.:tags
- list of tags to get the face of the entry.
<auto-tags-params>
is an alist with the following keys:
:max-level
- maximum level of the tree (default 2):source
- which feeds to use to build the tree. Can be:misc
(default) or(query . <query-params>)
.:original-order
- do not try to build a more concise tree by putting the most frequent tags closer to the root of the tree.:faces
- list of faces for groups.
<tag-group-params>
is an alist with the following keys:
:source
- which feeds to use to build the tree. Can be:misc
(default) or(query . <query-params>)
.:repeat-feeds
- allow feeds to repeat. Otherwise, each feed is assigned to group with the least amount of members.:face
- face for groups.
Available special forms:
:misc
- print out feeds, not found by any query above.
Also keep in mind that '(key . ((values)))
is the same as '(key (values))
. This helps to shorten the form in many cases.
Also, this variable is not validated by any means, so wrong values can produce somewhat cryptic errors. Sorry about that.
Example
Here is an excerpt from my configuration that was used to produce this screenshot:
(setq elfeed-summary-settings
'((group (:title . "GitHub")
(:elements
(query . (url . "SqrtMinusOne.private.atom"))
(group . ((:title . "Guix packages")
(:elements
(query . (and github guix_packages)))
(:hide t)))))
(group (:title . "Blogs [Software]")
(:elements
(query . software_blogs)))
(group (:title . "Blogs [People]")
(:elements
(query . (and blogs people (not emacs)))
(group (:title . "Emacs")
(:elements
(query . (and blogs people emacs))))))
(group (:title . "Podcasts")
(:elements
(query . podcasts)))
(group (:title . "Videos")
(:elements
(group
(:title . "Music")
(:elements
(query . (and videos music))))
(group
(:title . "Tech")
(:elements
(query . (and videos tech))))
(group
(:title . "History")
(:elements
(query . (and videos history))))
;; ...
))
;; ...
(group (:title . "Miscellaneous")
(:elements
(group
(:title . "Searches")
(:elements
(search
(:filter . "@6-months-ago sqrtminusone")
(:title . "About me"))
(search
(:filter . "+later")
(:title . "Check later"))))
(group
(:title . "Ungrouped")
(:elements :misc))))))
Automatic generation of groups
auto-tags
As described in the tree configuration section, there are two ways to avoid defining all the relevant groups manually, auto-tags
and tag-groups
. Both use tags that are defined in elfeed-feeds
.
auto-tags
tries to build the most concise tree from these tags. E.g. if we have feeds:
feed1 tag1 tag2
feed2 tag1 tag2
feed3 tag1 tag3
feed4 tag1 tag3
It will create the following tree:
- tag1
- tag2
- feed1
- feed2
- tag3
- feed3
- feed4
- tag2
The tree is truncated by :max-level
, which is 2 by default.
If tags don’t form this kind of hierarchy in elfeed-feeds
, the algorithm will still try to build the most “optimal” tree, where the most frequent tags are on the top.
To avoid that you can set (:original-order . t)
, in which case each feed will be placed at the path tag1 tag2 ... tagN feed
, where the order of tags is the same as in elfeed-feeds
. By the way, this allows reproducing the hierarchy of elfeed-org, e.g. this structure:
* tag1 :tag1:
** feed1
** feed2 :tag2:
** feed3 :tag2:
* tag3 :tag3:
** feed4 :tag2:
** feed5 :tag2:
** feed6 :tag2:
Will be converted to this:
- tag1
- feed1
- tag2
- feed2
- feed3
- tag3
- tag2
- feed4
- feed5
- feed6
- tag2
Whereas without :original-order
the structure will be:
- tag1
- feed1
- tag2
- tag1
- feed2
- feed3
- tag3
- feed4
- feed5
- feed6
- tag1
tag-groups
The second option is tag-groups
, which creates a group for each tag.
By default, each feed is assigned to its less frequent tag. This can be turned off by setting (:repeat-feeds . t)
.
E.g., the elfeed-org setup from the section above will be converted to this structure:
- tag1
- feed1
- feed2
- feed3
- tag3
- feed4
- feed5
- feed6
And with :repeat-feeds
:
- tag1
- feed1
- feed2
- feed3
- tag2
- feed2
- feed3
- feed4
- feed5
- feed6
- tag3
- feed4
- feed5
- feed6
Common options
Both auto-tags
and tag-groups
allow setting the :search
parameter.
The default value is (:search . :misc)
, i.e. use feeds that weren’t found by other queries.
Passing (:search . (query . <query-params>))
is another option.
Faces
Group faces by default use the elfeed-summary-group-faces
variable, which serves as a list of faces for each level of the tree. Individual group faces can be overridden with the :face
attribute.
Feed faces by default reuse the existing elfeed mechanism. The tags for feeds are taken from the elfeed-feeds
variable; if a feed has at least one unread entry, the unread tag is added to the list. This can be overridden by setting the elfeed-summary-feed-face-fn
variable.
Searches are mostly the same as feeds, but tags for the search are taken from the :tags
attribute. This also can be overridden with elfeed-summary-search-face-fn
variable.
Opening elfeed-search
in other window
If you set:
(setq elfeed-summary-other-window t)
Then RET
and M-RET
in the elfeed-summary
buffer will open the search buffer in other window.
elfeed-summary-width
regulates the width of the remaining summary window in this case. It is useful because the data in the search buffer is generally wider than in the summary buffer. The variable can also be set to nil
to disable this behavior.
Skipping feeds
tt-rss has a feature to disable updating a particular feed but keep it in the feed list. I also want that for elfeed.
To use that, set elfeed-summary-skip-sync-tag
to some value:
(setq elfeed-summary-skip-sync-tag 'skip)
And tag the feeds you want to skip with this tag. Then, running M-x elfeed-summary-update
will skip them. This won’t affect M-x elfeed-update
unless you:
(advice-add #'elfeed-update :override #'elfeed-summary-update)
Also watch out if you use elfeed-org and want to use the ignore
tag, because this package omits feeds with this tag altogether (configurable by rmh-elfeed-org-ignore-tag
).
Other options
Also take a look at M-x customize-group elfeed-summary
for the rest of available options.
Ideas and alternatives
The default interface of elfeed is just a list of all entries. Naturally, it gets hard to navigate when there are a lot of sources with varying frequencies of posts.
Elfeed itself provides one solution, which is using bookmarks to save individual searches. This can work, but it can be somewhat cumbersome.
elfeed-score is another solution, which introduces scoring rules for entries. Thus, with proper rules set, the most important entries should be on the top of the list. You can take a look at this video by John Kitchin to see how this can work.
However, I mostly had elfeed-score
to group entries to sets with equal scores, and I then processed one such set or the other. This is why I decided this package is a better fit for my workflow.
Another idea I used often before that is this function:
(defun my/elfeed-search-filter-source (entry)
"Filter elfeed search buffer by the feed under the cursor."
(interactive (list (elfeed-search-selected :ignore-region)))
(when (elfeed-entry-p entry)
(elfeed-search-set-filter
(concat
"@6-months-ago "
"+unread "
"="
(replace-regexp-in-string
(rx "?" (* not-newline) eos)
""
(elfeed-feed-url (elfeed-entry-feed entry)))))))
I’ve bound it to o
, so I would open elfeed
, press o
, and only see unread entries from a particular feed. Then I cleaned the filter and switched to the next feed. Once again, a tree with feeds is obviously a better tool for such a workflow.
The last solution I want to mention is elfeed-dashboard, although I didn’t test this one. It looks similar to this package but seems to require much more fine-tuning, for instance, it doesn’t allow to list all the feeds with a certain tag in a group.