Changing to iso 1806 style dates cuz who needs anything else amiright?
DISTINCT ON with ORDER BY acts acts like window functions with ranking/row_number. https://www.youtube.com/watch?v=XuGxGP8quMQ
-- Set the variable "query" to the contents of the sql file
\set query `cat my_rad_query.sql`
-- Use the variable as the query and copy to csv
copy (:query) TO STDOUT WITH CSV HEADER \g 'my_rad_query_result.csv'
cli benchmarking tool https://github.com/sharkdp/hyperfine
yay -S hfsprogs
mkdir gumption-5
sudo mount -t hfsplus -o force,rw /dev/sda2 ~/gumption-5
sda2 found from lsblk -f
tldr don't loop
something | xargs -I % sh -c 'command where every occurrence of % is replaced by one of the args from xargs'
echo -e 'meow1\nmeow2\nmeow3\n' | xargs -I % echo "any little % is the spawn of %'s parent"
echo -e 'this\nthat\nthe other\n' | xargs -I % sh -c 'echo %; echo %'
c/o dear Travitz
To get a video to work with WhatsApp it needs to be MP4 (H264 + aac). From Quora of all places!
ffmpeg -i file-from-guvcviewer.mkv -vcodec libx264 -acodec aac functioning-whatsapp-vid.mp4
git ls-files --others --exclude-standard
Add something to line ending with ...
'<,'>g/\.\.\.$/norm A something
Works with v too!
Suppose you've got:
| header1 | header 2 |
| ------- | -------- |
| a | c |
| b | b |
You can sort by header 2 via sort -t'|' -k3... the left bar counts as one, then add for columns. Just do column # plus 1. I used it in vim via visual selection like:
:'<,'>.! sort -t'|' -k3
From Christian Tietze post
Get an overall diff from start to finish on a branch w/o remote compares "beginning of time" with the tip
git diff --name-only $(git rev-list --max-parents=0 HEAD) HEAD
mkdir ~/<drive-name-whatevs-shrug>
Find the drive via lsblk -f
sudo mount -t hfsplus -o force,rw /dev/sdXY ^^^^dir-above
udisksctl power-off -b ^^^^dir-above
sudo eject/dev/sdXY
From -> blog post
SELECT
pid,
now() - pg_stat_activity.query_start AS duration,
query,
state
FROM pg_stat_activity
WHERE (now() - pg_stat_activity.query_start) > interval '5 minutes';
This does a 3 second slide delay
feh -rF -D 3 <folder>
This allows the use of non-escaped parens and | for composing a single g delete command with multiple search terms
g/\v(\d\d:\d\d|^$)/d
This applied to a visual range (could use other ranges also) along with an awk shell out cleans up a copy pasta from codeship.
!awk '{print $1" "$2}'
python -c 'import sys, yaml, json; yaml.safe_dump(json.load(sys.stdin), sys.stdout, default_flow_style=False)' < in-file.json > out-file.yaml
also %s/\(\<..$\)/US\1... pretty cool, adds US to someting like area: KY --> area: USKY
Used this to grab the filename for a hard to autocomplete migration file.
r! echo %
A little better 🤷?
:exe ":silent ! echo % | xclip -sel clip" | redraw!
To format json:
nnoremap <leader>j :%!jq ''<CR>
To format html:
nnoremap <leader>h :%!xmllint --format --encode UTF-8 --html -<CR>
This takes a visual selection from vim and hands it off to awk for 'filtering'.
:'<,'>! awk 'match($0, /^.+\. (.*)$/, a) {print NR". " a[1]}'
'<,'>! awk 'match($0, /^[^.]+\. (.*)/, a) {print NR". " a[1]}'
The Awk match function takes a string, a pattern, and a variable name for the resultant array of matches. In this case a[0] would be the whole line and a[1] the one and only capture group from the pattern.
This specifically matches something up to a literal period character followd by a space (the existing numbers), captures everything after that space, and then re-adds the line numbers to the selection based on Awk's built-in NR variable which is the 'number of records so far' starting from zero.
As a personal note, this solution represents a continued resistence to my learning proper vimscript functions. I've looked into some plugins and have a very vague idea of what functions look like, what they can do etc, but there has been a hurdle to making my own that I've been unwilling or unable to cross thus far. In defense I can say that using Awk bolsters familiarity with another Unix tool that can be used outside of Vim, and that this specific problem is an awky/sedy type line problem for which these programs excel.
git status --porcelain | awk '{print $2}' | xargs rm
~/.dropbox-dist/dropboxd &
for dir in $(ls -t | tail -n +2); do rm -rf $dir; done
The ls -t lists files/directories by most recent. tail -n +2 lists output starting with the second line.
I can't believe I didn't put this in here before (had encountered this problem and solution once already). If you're using / in a recursive macro, if you continue to have matches after the 'first round' you'll keep running through the file doing more action than you had anticipated and it will do wild things until it runs out of search matches. This is because by default vim has a boolean variable called wrapscan set to true. You can turn this feature off with :set nowrapscan and the search will end with the end of the file.
Please stop doing Capybara.current_session.driver. You'll have to kill the pry session because the object is so big that it'll scroll seemingly forever. There are more monster objects as well, if I recall a request is similarly large and "scrolly".
I knew you could do something like :g/pry/d to delete all lines with pry in them. I use this to clean up from heavy debugging sessions in both js and rb files.
Today I used '<,'>g/do/:norm 0$gEa, data: { action: 'header#linkTest' } do add a bit of text to every line that contained /do/ within a visual selection.
..... time elapses ....
In the aforementioned snippet, I used CTRL-r j to dump the contents of my macro onto the command line. The contents of j were: 0$gEa, data: { action: 'header#linkTest' }. Turns out from reading more of Power of G I could have just done something like:
:g/pattern/normal @q
Occassionally I find myself wanting to put things side by side in the same file in vim. I've used this to easily compare similar parts of Rails schema.rb. While I'd rather not sort the lines in the schema it's easy enough to copy paste the comparison targets and put them side by side in a scratch file. The technique is outlined here and I've forgotten and gone back enough times it merits an entry here!
If you've got the contents butted up against the left side go to the block (the bit you want as a 'right column'). Ctrl-v to start VISUAL BLOCK mode, go to bottom of the selection, $ to the end and d. Move to the 'left column', hit A on the first line and space out until you've cleared the longest line of 'left column'. <esc> to normal mode and p to paste the select. The [link to SO] (https://stackoverflow.com/a/27542895/5889617) has a nice gif show-and-tell if this didn't make sense.
It's a good idea if you're doing line-wise comparisons to sort both the 'left' and 'right' columns by using V to select the lines followed by :sort or :sor. Visually at the vim display line it'll show like: :'<,'>sor. That's just your visual selection on the vim command-line.
Today I was curious how many places I might need to change a thing in trying to decide my course of action. The command is as follows:
ag -c 'embedded' | awk -F ':' '{n=n+$2}END{print n}'
ag -c PATTERN will give you output like:
path/to/file:3
path/to/file:15
path/to/file:1
The Awk bit breaks the line with a specific Field separator -F ':' and then accumulates the second field (the one with our find count) in the variable n. After Awk is done it will run anything in the END block, in this case just printing out our accumulated variable.
vim $(find app/views/purchase_applications/steps/ | awk '{ if (NR != 1) { print } }')
The awk bit removes the directory itself from the list of files.
It occured to me from the last section that what I really want is to multiline grep. Grep can't do that. Neither can ag. Thus the discovery of pcregrep which has a multiline option.
pcregrep -r -M 'hidden {\n display: none;\n}' app/assets/stylesheets/
Just for kicks, a fairly rediculous command to give me the actual number:
pcregrep -c --files-with-matches -r -M 'hidden {\n display: none;\n}' app/assets/stylesheets/ | awk -F ':' '{ n = n + $2; } END { print n }'
-> 13
Lolz.
You can use -- after some options to a command to explicitly state that you're done with options and that everything after should be argument.
I used this to learn how many times we've duplicated a style that toggles visibility alone (display: none). We use some version of BEM for naming our styles so the command that was failing:
ag --sass '--hidden' <- Causes ag to go 'wait, what option is --hidden?'
The working one:
ag --sass -- --hidden
On 12/4/2018 I talked about the following command:
ag --haml -l 'placeholder' | xargs sed -i -e "s/placeholder: '\(.*\)'/placeholder: '\L\1'/g" -e "s/placeholder: '\(.*\)'/placeholder: '\u\1'/g"
This did what I thought I wanted it to, until yesterday when I found out that our javascripty mortgage calculators were broken. It turned out to be a lesson to be wary of using * in Sed replacements. * is greedy! On a handful of lines, those in the caclulator js files, the lines with the placeholder replacement also held javascript variables for stimulus that were case sensitive...
placeholder: 'Length of loan', data: { 'graph-input': 'loan-term', target: 'mortgage-payment.loanTerm',
became:
placeholder: 'Length of loan', data: { 'graph-input': 'loan-term', target: 'mortgage-payment.loanterm',
changing loanTerm to loanterm and causing stimulus to throw a missing target element error and kill js execution. That greedy * ran the sed lower-case and capital-case trickery all the way to the last single quote on the line!
So..... beware the greedy star!
Xargs can take -r or --no-run-if-empty to avoid failure in a script. I made a little script to remove all curly quote characters, change ndash to mdash, and remove empty whitespace:
#! /bin/bash
set -xe
ag -l '[[:space:]]*$' | xargs -r sed -i 's/[[:space:]]*$//'
ag -l '’' | xargs -r sed -i "s/’/'/g"
ag -l '“' | xargs -r sed -i 's/“/"/g'
ag -l '”' | xargs -r sed -i 's/”/"/g'
ag -l '–' | xargs -r sed -i 's/–/—/g'
Without the -r, if the project doesn't have results from the ag, sed will throw sed: no input files.
note: the set -xe shows the command run to stdout (x) and exits the whole script if a line fails (e).
ag --haml -l 'placeholder' | xargs sed -i -e "s/placeholder: '\(.*\)'/placeholder: '\L\1'/g" -e "s/placeholder: '\(.*\)'/placeholder: '\u\1'/g"
Goes through and lowercases the submatch (the placeholder text), on a second pass capitalizes the submatch.
example:
.* placeholder: 'My Field'.* -> .* placeholder: 'My field'.*
Sed special backslash sequences
Another more complicated one:
ag --haml -l '%span\.form' | xargs sed -i -e 's/%span\.form\(.*\)field \(.*\)/%span.form\1field \L\2/g' -e 's/%span\.form\(.*\)field \(.*\)/%span.form\1field \u\2/g'
example:
.*%span.form__label-radio-field Old Password -> .*%span.form__label-radio-field Old password,
where there can be different class text between form and field.
- Copy Pasta the text into vim.
:set textwidth=80gggqG
the magic here is gq.
git diff master...HEAD
or
git diff master..
In the past I've had problems trying to create a recursive macro and it turns out I'm just a dum dum. If you're creating a macro and you want to call that same macro from within itself during that very definition, the register itself needs to be empty first. I believe you can clear it with:
let @j=''
I've got a habit of using the same couple of registers for my macros, namely j and k because they're nearby. Pasting the contents of the register contianing the macro into a file and then altering the macro to add @register-in-question works as long as you've got a starting macro that doesn't end with a special character. i.e. when I want to expand a line with entities separated by spaces into having each entity on its own line (putting hash k/vs in a single column to sort maybe) I'd have:
f s^M^C^C
(since I'm in the habit of using ctrl-c to exit insert mode) but when I paste that into a file I've got 'f s' and have lost the special characters. My past problems relate to already having something stored in the macro and thus during recording it kicks off the previous version and wreaks havoc. Back to having some macros be one shot wonders.
..... ..... time elapses .... ....
hold the phones, a way around this is to use the vim command-line, which retains special characters. Go into command mode, start out using the let syntax to redefine the register, ctrl-r + register, ctrl-e (if you've got emacs style shortcuts defined) and then add the @register to the end. wooohooo. (looking back at the thoughtbot article where I first saw the paste trick I see that the let style was described also... with a repetitive ctrl-r and without the bit about special characters.
:%bd|e#
Sort of a workaround because it actually closes all, then reopens the last one. The alternatives seemed arduous.
From Solomon's comment on Skorks shortcuts blog post
:cq
This quits vim with an error code so that whatever dire consequence may be avoided. For example if you're rebasing and realize you shouldn't be, or if you've ctrl-x ctrl-e and changed your mind about firing the command.
From this stackoverflow post and leading to the aforementioned Skorks blog post are more complete lists of all the readline shortcuts including the new-and-fantastic (to me) alt word based movements. This has prompted me to take a look at redoing my keymap to drop left or right shift in favor of positioning alt in a friendlier place. tbd.
I wanted to find all js files in our project but exclude those in the vendor directory. This SO post describes many ways of doing it. I found the -prune option to be confusing. Further down the clear winner by 1518 votes describes using -not -path and so my final command:
find app/javascript '*.js' -type f -not -path '*vendor*'
The other option that was working intially is to ag or grep the results accordingly ag -v vendor to remove the vendor files from the results of the find
vim $(find -iname '*.html' | xargs ag -Ql 'w[l]')
Passing multiple files to vim with the $() instead of piping things into vim via xargs wont wonk the terminal display as mentioned on 4/20. After the files are in vim's buffer, you can record a macro including the :wn to save the file and move to the next file in the buffer (As seen here. In that next file I stopped the macro, opened a newline and pasted from the macro register to add the recursive invocation piece to the end blahblahkkkddd/meowIetc^C^C:wn^M@j (I always record to 'j', cuz its under my finger) (I wasn't able to just add the @j while I was recording the macro, something lacking in my understanding there). Yank that line back into register j, then @j to kick off the recursion! video example.
osx:
osascript -e 'display notification "Notification text" with title "Title"'
linux (ubuntu):
notify-send <message>
In conjuction with & at the end of a command (sending a long running job to the background), these will alert you as to when the job might be finished.
You can ls <options> | xargs vim to dump a load of files into vim.
:args shows all the filenames (messy).
:n goes to the next file and if you edit you can smash the commands together :wn
:prev goes to the previous file in the list.
It seems like all those arguments are dumped into :buffers (much prettier list of the same files as :args) so I was able to use my custom 'cruise the buffers' key bindings. I just cruised the whole list and :bn seems to satisfy the same criteria as :n in the sense that after you've reached the end of the file set you're able to quit without forcing the issue (no E173).
A big caveat is that terminal rendering gets messed up after you exit vim... in alacritty with tmux, in plain alacritty, and in gnome-terminal...
grep -rl --include *.svg <search pattern>
I used this with giving vim multiple files.
When I ssh into a qa server, I only get one tmux pane (apprently default ssh allows one connection) unless I fiddle with some ssh config. Saying I only have one, and I was going to be in rails console anyway, I can run
Rails.application.load_tasks
followed by
Rake::Task['some_task_name'].invoke
to run the task from the console. If there is a namespace it works the same so Rake::Task['session:log_from_data'].invoke is valid.
Also you can list tasks via Rake::Task.tasks
delete blank lines: :g/^\s*$/d
delete whitespace: :%s/\s\+//g
<\([^<]*\)> will match all fully formed html/xml tags, anything in < >. I've found this useful for scraping bits of websites by copying the outer html of an element in the browser inspector and dumping it in vim. With this I can delete all the tags in a single substitue command :%s/<\([^<]*\)>//g
%s/@include justify-content(\(.*\));/justify-content: \1;/g
From SO post here.
Example case: rebase file squashing commits. Changing pick to s on a bunch of lines. I used to have absolute and relative line numbering in vim but on pairing with rtravitz and my lead its much easier to reference lines in absolute. With relative line numbering it was possible to record the macro like ciws^C^Cj where each run of the macro bumps down a line after changing the word, and play it to the end of the commits to be squashed. With absolute line numbering this isn't possible without doing math in my head. nope. so:
Record macro making sure that it doesn't move from one line (this is contra my previous method). In the example case it would be ciws^C^C. Visual Line select the remaining lines and hit :. This produces the command line with selection: :'<,'>. Add the macro play like: :'<,'>norm @j, and there you have it.
In the SO post someone also showed a capture group example that swaps two words separated by a comma. This pertains to an earlier diary post about sed (2/27/2018).
'<,'>s/\(\w\+\), \(\w\+\)/\2, \1/
Let me separate that a bit
'<,'> visual range
s/ substitue
\(\w\+\), \(\w\+\) pattern with the capture groups, first becoming \1 and second becoming \2
/ pattern/replacement separator
\2, \1 the replacement string utilizing the capture groups
/ end of subsitute phrase
The keybindings from vim-tmux-navigator overshadow the default tmux selection mode vi keybindings like here. I ran across a post by Chris Toomey that remaps the start selection keys to mimic vim here. Seems to work.... a little twitchy. Maybe an Alacritty thing? It doesn't seem to work at all in gnome-terminal.
Need to replace all instances of sass-flex mixins with regular declarations that in compilation get expanded to the browser flag editions by way of autoprefixer.
ag -Ql "@include justify-content" | xargs sed -i 's/@include justify-content(\([^\)]*\));/justify-content: \1;/'
The escaped parens in the matching expression \( and \) in \([^\)]*\) create a variable \1 that is the contents between the unescaped parens in justify-content(some variable stuff in here). This allows all the different values inside to be output in the replace expression
I've come across perhaps a more general way:
ag -Ql "@include transform" | xargs sed -rn 's/@include transform\((.*?)\);/transform: \1;/'
This will handle something like
@include transform(translate(-50%, -50%) skewY(-4.2deg));
I want to test my sed string on a file, or many files before doing something wild.
sed -n "s/search-term/replacement/p" file-name
Then, when its time to alter sed -i "s/search-term/replacement/" file-name
I also can do a count on occurances across a directory to compare after the change
ag -Q "search-term" | wc -l
I've got a dotfiles folder living in the git dir where I keep my dotfiles for git control purposes.
move the dotfile you want (I'll be referring to my i3 config file):
mv ~/.config/i3/config ~/git/dotfiles/config
symlink the dotfile so it can be loaded:
ln -s ~/git/dotfiles/config ~/.config/i3/config
I always forget the syntax: ln -s followed by <where the file actually lives> followed by <where I want to file to pretend to live>
ctrl-x ctrl-e
fc
find . -type f | xargs sed -i "s/#{phone}/#{link_to phone, \"tel:#{phone_raw}\", class: 'active-states-state__link'}g"