To install this package, run in Emacs:
M-x package-install RET vlf RET
Emacs minor mode that allows viewing, editing, searching and comparing large files in batches, trading memory for processor time. Batch size can be adjusted on the fly and bounds the memory that is to be used for operations on the file. This way multiple large files (like terabytes or whatever) can be instantly and simultaneously accessed without swapping and degraded performance.
This is development version of the GNU ELPA VLF package. Here's what it offers in a nutshell:
GNU Emacs 23 and 24 are supported.
M-x vlf PATH-TO-FILE
Emacs' Unicode support is leveraged so you'll not see bare bytes but characters decoded as if file is normally opened. This holds for editing, search, indexing and comparison.
Regular Emacs integers are used, so if you use 32-bit Emacs without bignum support, VLF will not work with files over 512 MB (maximum integer value).
To have vlf offered as choice when opening large files:
You can control when vlf-mode is invoked or offered with the vlf-application customization option. By default it will offer VLF when opening large files. There are also options to never use it (you can still call vlf command explicitly); to use it without asking for large files or to invoke it on all files. Here's example setup such that vlf-mode automatically launches for large files:
(custom-set-variables '(vlf-application 'dont-ask))
To disable automatic usage of VLF for a major mode, add it to vlf-forbidden-modes-list.
To disable automatic usage of VLF for a function, for example named func defined in file file.el:
(vlf-disable-for-function func "file")
All VLF operations are grouped under the C-c C-v prefix by default. Here's example how to add another prefix (C-x v):
(eval-after-load "vlf" '(define-key vlf-prefix-map "\C-xv" vlf-mode-map))
To see which part of the file is currently visited and how many batches there are in overall (using the current batch size), look at the VLF section in the mode line, file size is also there.
By default VLF gathers statistics over how primitive operations perform over file and gradually adjusts batch size for better user experience. Operations involving multiple batches are tuned more adventurously. Overall the more jumping around, searching, indexing, the better performance should get.
The vlf-tune-max option specifies maximum size in bytes a batch could eventually get while tuning.
Profiling and tuning can be disabled by:
(custom-set-variables '(vlf-tune-enabled nil))
Or set vlf-tune-enabled to 'stats to profile but not change batch size.
Use M-x vlf-set-batch-size to change batch size and update chunk immediately. Default size offered is the best according to tune statistics so far.
C-c C-v + and C-c C-v - control current batch size by factors of 2.
Scrolling automatically triggers move to previous or next chunk at the beginning or end respectively of the current one.
C-c C-v n and C-c C-v p move batch by batch. With positive prefix argument they move prefix number of batches. With negative - append prefix number of batches.
C-c C-v SPC displays batch starting from current point.
C-c C-v [ and C-c C-v ] take you to the beginning and end of file respectively.
C-c C-v j jumps to particular batch number.
Continuous chunk recenter around point in current buffer can be toggled with C-c C-v f.
C-c C-v s and C-c C-v r search forward and backward respectively over the whole file, batch by batch. C-c C-v % does search and query replace saving intermediate changes.
C-c C-v o builds index over whole file for given regular expression just like M-x occur. Note that even if you prematurely stop it with C-g, it will still show what's found so far.
Result buffer uses vlf-occur-mode which allows to optionally open new VLF buffer on jump to match (using C-u before hitting RET or o), thus having multiple simultaneous views of the same file. Also results can be serialized to file for later reuse.
C-c C-v l jumps to given line in file. With negative argument, lines are counted from the end of file.
If editing doesn't change size of the chunk, only this chunk is saved. Otherwise the remaining part of the file is adjusted batch by batch. vlf-save-in-place customization option controls if temporary file should be used in such case.
Use M-x vlf-ediff-files and M-x vlf-ediff-buffers to compare files/buffers batch by batch (batch size is queried in case of files or taken from the first buffer in case of buffers). Moving after the last difference in current chunk searches for following one with difference. The other way around if looking for difference before the first one.
A couple of hooks are run whenever updating chunk: vlf-before-chunk-update-hook and vlf-after-chunk-update-hook.
Some operations may trigger multiple chunk moves. There are a couple of hooks that run in such cases: vlf-before-batch-functions and vlf-after-batch-functions. They are passed one argument which specifies type of operation that runs. Possible values are the symbols: write, ediff, occur, search and goto-line.