Logview mode

Logview major mode for Emacs provides syntax highlighting, filtering and other features for various log files. The main target are files similar to ones generated by Log4j, Logback and other Java logging libraries, but there is really nothing Java-specific in the mode and it should work just fine with any log that follows similar structure, probably after some configuration.

The mode is meant to be operated in read-only buffer, so all the command bindings lack modifiers.

Out-of-the-box the mode should be able to parse standard SLF4J (Log4j, Logback) files as long as they use ISO 8601 timestamps, Apache error logs, PHP Monolog logs, and certain UNIX files in /var/log.

Installation

Logview is available from MELPA (both stable and unstable). Assuming your package-archives lists MELPA, just type

M-x package-install RET logview RET

to install it.

Alternatively, installing Logview from source is not difficult either. First, clone the source code:

$ cd SOME-PATH
$ git clone https://github.com/doublep/logview.git

Now, from Emacs execute:

M-x package-install-file RET SOME-PATH/logview

Alternatively to the second step, add this to your .emacs file:

(add-to-list 'load-path "SOME-PATH/logview")
(require 'logview)

Submodes

Since there is no standard log file format, Logview mode has to try and guess how the log file it operates on is formatted. It does so by trying to parse the first few lines of the file against various submodes it has defined.

If it succeeds in guessing, you will see major mode specifed as ‘Logview/...’ in the modeline, where the second part is the submode name.

In case it fails, you will see it complain in the echo area. The buffer will also not be highlighted or switched to read-only mode, so you will be able to edit it.

What to do if Logview mode fails to guess format

This depends on why it fails. If the mode doesn’t know about this log format at all, customize the relevant options. C-c C-s will show you the three variables that are important for submode guessing. You will need to customize at least one of those, or maybe all three. All the variables are well-documented in customization interface.

However, it is also possible that Logview fails because it looks only at the first lines of the buffer. As of 0.14 this can be up to 500 lines (see option logview-guess-lines), but in the unlikely case real log entries in your files start even later, you can customize this value. However, there is an optimization in Logview: if it discovers several lines that do look like log entry start, yet cannot be understood by the mode, guessing is aborted. This is done because otherwise guessing could take very long time and would still be unsuccessful. This optimization is triggered upon seeing logview-max-promising-lines such lines (default value is just 3). You can always customize this setting if needed, but remember, that this can lead to very long guessing times.

Finally, you can always force Logview to switch to appropriate submode using C-c C-c shortcut. Remember that you still need to have it registered first. In case you don’t want to manually select submode each time in such cases, you can set up some automatic switching from a hook using Elisp function logview-choose-submode.

If you think your log format is standard enough, you can open an issue and request format addition to the list of mode built-ins.

Commands

Nearly all commands have some use for prefix argument. It can be usually just guessed, but you can always check individual command documentation within Emacs.

When buffer is switched to read-write mode, Logview automatically deactivates all its commands so as to not interfere with editing. Once you switch the buffer back to read-only mode, commands will be active again.

Movement

• All standard Emacs commands
• Move to the beginning of entry’s message: TAB
• Move to next / previous entry: n / p
• Move to next / previous ‘as important’ [*] entry: N / P
• Move to next / previous entry in the navigation view: M-n / M-p
• Move to the next / previous entry with large timestamp gap after the previous: z n / z p
• Same as above, but only considering entries in the same thread: z N / z P
• Move to first / last entry: < / >

[*] ‘As important’ means entries with the same or higher level. For example, if the current entry is a warning, ‘as important’ include errors and warnings.

Narrowing and widening

• Narrow from / up to the current entry: [ / ]
• Widen: w
• Widen upwards / downwards only: { / }

Filtering by entry level

• Show only errors: l 1 or l e
• Show errors and warnings: l 2 or l w
• Show errors, warnings and information: l 3 or l i
• Show all levels except trace: l 4 or l d
• Show entries of all levels: l 5 or l t
• Show entries ‘as important’ as the current one: + or l +

Always show entries of certain levels

It is possible to always display entries of certain levels, regardless of any additional text filters.

• Always show errors: L 1 or L e
• Always show errors and warnings: L 2 or L w
• Always show errors, warnings and information: L 3 or L i
• Always show all levels except trace: L 4 or L d
• Disable ‘always show’ feature: L L or L 0

Filtering by entry’s logger name, thread or message

See more detailed description below.

• Edit current name, thread and message filters: f (pops up a separate buffer)
• Add name include / exclude filter: a / A
• Add thread include / exclude filter: t / T
• Add message include / exclude filter: m / M

Resetting filters

• Reset level filter: r l
• Reset name filters: r a
• Reset thread filters: r t
• Reset message filters: r m
• Reset all filters: R
• Reset all filters, widen and show all explicitly hidden entries: r e

Views

See more detailed description below.

• Switch to a view: v
• Choose navigation view (for M-n and M-p): V n
• Select a view to highlight its entries: V h
• Remove view highlighting: V u
• Save the current filters as a view for this submode: V s
• Save the current filters as a global view: V S
• Edit submode views: V e (pops up a separate buffer)
• Edit all views: V E (pops up a separate buffer)
• Assign a quick access index to the current view: V i
• Delete a view by name: V d

You can also switch to views using their quick access index: M-0..M-9 or e.g. 1 4 v (for quick access index 14). Prefix argument works also for V n and V h.

Explicitly hide or show individual entries

• Hide one entry: h
• Hide entries in the region: H
• Show some explicitly hidden entries: s
• Show explicitly hidden entries in the region: S
• Show all manually hidden entries in the buffer: r h

In Transient Mark mode h and s operate on region when mark is active.

Entry timestamp commands

• Replace timestamps with their difference to that of the current entry: z a.
• Same as above, but only within the same thread: z t
• Go to the entry difference to which timestamp is shown: z z
• Don’t show timestamp differences: z A
• Don’t show timestamp differences for this thread: z T

Timestamp differences are displayed in seconds.

Explicitly hide or show details of individual entries

The mode terms all message lines after the first ‘details’. Oftentimes these contain exception stacktrace, but most logging libraries let you write anything here.

• Toggle details of the current entry: d
• Toggle details of all entries in the region: D
• Toggle details in the whole buffer: e
• Show all manually hidden entry details in the buffer: r h

In Transient Mark mode d operates on region when mark is active.

Change options for the current buffer

These options can be customized globally and additionally temporarily changed in each individual buffer.

• Change gap length for z n and similar commands: o g or z g
• Toggle Auto-Revert mode: o r
• Toggle Auto-Revert Tail mode: o t
• Toggle ‘copy only visible text’: o v
• Toggle ‘search only in messages’: o m
• Toggle ‘show ellipses’: o e

Miscellaneous

• Pulse (briefly highlight) the current log entry: SPC
• Manually choose appropriate submode and timestamp format: o s or C-c C-c
• Customize options that affect submode selection: o S or C-c C-s
• Bury buffer: q
• Refresh the buffer (appending, if possible) preserving active filters: g
• Append log file tail to the buffer: x
• Revert the buffer preserving active filters: X
• Universal prefix commands are bound without modifiers: u, -, 0..9

Filters

In addition to level filtering, Logview supports filtering by entry’s logger name, thread and message.

These filters are regular expression and come in two kinds: include and exclude filters. If at least one include filter is set, only those entries where relevant part matches at least one of regular expressions are shown, all others are filtered out. If any exclude filter is set, all entries where relevant part matches its regular expression are filtered out (regardless of any other filters) and hidden.

Easiest way to add filters is by using a / A commands (add include/exclude name filter correspondingly), t / T (thread filters), and m / M (message filters). You can reset all filters of given type: r a for name, r t for thread and r m for message filters, R for all filters at once.

However, oftentimes you need to adjust existing filters, e.g. to fix a typo or simply change one, while keeping others the same. For this purpose use f command. It pops up a separate buffer with all currently active filters (both level and text), which you can edit as you like, using any standard Emacs features.

Lines beginning with ‘#’ character are comments. They and empty lines are ignored. Lines for level filters must start with prefix ‘lv ’ (note the single trailing space!) for normal filtering or ‘LV ’ for ‘always show’ pseudo-filter, and contain the textual level representation afterwards, just as you would see it in the logfile. For example:

lv DEBUG
LV ERROR

means “show all entries except trace-level, and additionally show all errors with no regard to text filters.”

Lines for text filters are similar. They must start with ‘a+ ’ (again, remember the trailing space) for name inclusion filters, ‘a- ’ for name exclusion, and similarly ‘t+ ’, ‘t- ’, ‘m+ ’, ‘m- ’ for thread and message filters. Additionally, multiline message filters must use ‘.. ’ prefix (two dots and a space) for continuation lines. For example:

a+ database
m+ ^insert.+
.. values

means “show entries with word ‘database’ in the logger name and which message has a line beginning with ‘insert’ and the line after that beginning with ‘values’.”

The buffer mode has some syntax highlighting support, so you should see if anything goes wrong. The easiest way to figure it out is to add a few filters using commands described earlier and then open this buffer with f and see how they are represented.

Filter regexp details

Regular expressions can be matched against entry parts either case-sensitively or case-insensitively, depending on standard Emacs variable case-fold-search.

Filters are matched against relevant entry parts as strings, not against the whole buffer. Therefore, you can use ^ and $ special characters for the expected meaning. For example, adding ^org as name exclusion filter will hide all entries where logger name begins with string ‘org’.

Unlike name and thread filters, message filters can span multiple lines. To enter linefeed in message buffer (after m or M) use C-q C-j. When editing a multiline filter with f, prefix all continuation lines with ‘.. ’.

Commands a, A, t and T default to the name (or thread) of the current entry. You can also use C-p (<up>) to browse history of previously entered values and C-n (<down>) for a few default values.

Views

Views are named sets of filters that you can activate quickly. They are especially useful if you use name or message filters a lot, and often find yourself typing in the same filters over and over again.

The easiest way to define a view is by first adding all the filters you need. This way you can see in the buffer if the filtering result matches what you expect. After you are satisfied, type V s and a name for the new view. Notice that the mode line now displays name of the view in square brackets after the submode name, e.g.:

Logview/SLF4J [useful-view-1]

Now type R to reset all the filters. All previously hidden entries will be shown again and the view name disappear from the mode line. However, to restore the filters now you don’t have to re-create them one-by-one. Simply type v and whatever name you used when saving your first view. You can also use text completion to pick among all the defined views.

To make choosing views even easier, you can optionally assign quick access indices to views. For this, activate a view normally (or have it just saved), type V i and enter a number, say 3. After this, the view can be quickly activated again by typing M-3 or 3 v.

Remember that further filtering doesn’t affect view definition. If you want to change a view, save filters as a view with the same name again, and confirm that you do want to replace the previous definition. Alternative way is to edit views using V e. This pops up a separate buffer just like f command does, but instead of filters you will edit all defined views for the current submode at once. This way you can change existing definitions, delete unneeded or add more. Commands like V s or V d (delete a view by name) can be seen as just a convenience.

Views come in two kinds: globally accessible and bound to a specific submode. This distinction is important if you use logs of different kinds. Most often you need submode-specific views, because text filters usually can’t be meaningfully applied without changes to different programs. When you use v command, only the views for the current submode plus any global views are available for selection.

In addition to applying view filters, it is also possible to move between entries in a view without activating it. For this, define a view and then set it as as a ‘navigation view’ with V n command. After this, use commands M-n and M-p to quickly navigate forward and backward. Remember that these commands skip all hidden entries, whether because of your main view (or filters) or manual entry hiding.

Finally, you can highlight all entries in a view, or, more precisely, those that are visible currently. This can be done with V h command. Cancel this by highlighting entries from a different view or removing highlighting altogether with V u.

To summarize:

• You can have any number of named views. Their definitions are stored permanently across Emacs session and are available from all Logview buffers.

• At any time you can switch to a view, i.e. replace current filters with those stored in the view’s definition. Changing filters itself doesn’t alter any view definitions.

• You can choose one navigation view, independently from the currently applied view. It is used by commands M-n and M-p.

• You can highlight entries of a view, again, independently from current or navigation view.