I'm new to zsh and I'm trying to figure out how to get tab completion to work so that when I typed part of the name of a file that's not in the current directory zsh will complete it.
The idea is that I have some scripts in ~ and c:\MyStuff\bin and I'd love for zsh to try and complete those (executable) scripts when I'm in other directories.
Being able to complete files that are anywhere in my path would be nice, but if it's easier to complete files using a list of directories set in my .zshrc, well, that would work fine too.
If anyone has any pointers for resources about how to do this, or even advice like "This will / won't work in zsh" that would be great. zsh seems open-ended enough that it ought to be able to do this AND ALSO I've searched long enough without finding anything that I wouldn't be surprised if there's nothing at the end of this rabbit hole :)
Thanks in advance!
So I decided to invest the time and actually try and read the (very, very thorough) zsh completion docs. There's a lot there, including a section which says that zsh actually does this out of the box:
8 Command Execution
"...the shell searches each element of $path for a directory containing an executable file by that name..."
Turns out this does work for me, it just takes a long time (more than a second or two) and so I thought it wasn't working.
Next up: looking at why it's taking so long - perhaps it's my very long $path variable :)
Suppose I want to access an online dictionary and need to look for a specific word. I just like to have the specific part of data, which is those related to word and its translation as input of AWK,any idea?
In other words, I just want to have on my machine a margin of data, How can I prevent downloading all the data and hopefully save space and time. Is there any way to do so without downloading all the data to local machine?
This question is related to my last question here.
Edit 1:
I select dictionary as an example because when you want to look up for a word, it is enough to access a specific part of data and there is no need to process whole of it.
I am not an expert in programming so i was thinking I can modify this answer to make it work(that is why I add AWK tag again). I dont use any specific OS or tool. this is just a basic idea to see what are the possibilities so I dont know how can I improve the tags.
awk cannot download. You must download the file and pipe it into a command that terminates as soon as it finds a result:
wget -qqO- http://example.com/path |grep -wim1 "word"
wget -qqO- URL will have no output other than the content of the given URL, which is placed on standard out so you can then parse it. grep -wim1 "word" will find the first bounded word matching "word" and then terminate. If you don't need it outputted, you can use -wiq instead. If the dictionary has one word per line (and nothing else), you're better off with -x instead of -w so that you can match "can" in its entirety rather than "can't" (' is a word boundary). Remove the -i if you want to match case.
In the comments, you asked:
it may improve to jumpt to start of "w" character maybe so not to download whole data from "a" to "w". is it possible? I guess not
Some programs can "resume" downloads and you may be able to play with that, but you'd have to guess where to start. This would be a lot of work and you might seek too far and therefore fail to get a match.
If you are querying this dictionary more than once, I'd recommend downloading it and saving it so you can query it locally. Even the largest dictionary I know of is only 213MB (compressed, search with zgrep), though I am assuming you're talking about a traditional word list rather than a hash table or other arbitrary data form. Of course, anything longer would take such a long time to download that you'd only want to do it once.
If you really don't want to store it locally, you should probably consider a database rather than a flat file.
The program that I am trying to run takes the form program_name --arg=/some/path/goes/here, but zsh cannot perform tab completion on that path when it is in the argument flag. I end up having to type progra<TAB> /so<TAB>/pa<TAB>... to complete the path and then go back to add the --arg= part of the command. Is there a more efficient way to go about doing this?
It's likely that zsh is still trying to do file completion in your situation, it's just considering the --arg= to be the initial part of the file name causing it to not find any matches. Since there aren't any universal rules for how commands will interpret their arguments the safest option is for the shell to not make any assumptions about those arguments.
Many commands which use that type of syntax will work the same if you use a space rather than the = sign, such as program_name --arg /some/path/goes/here. Since that would be a separate argument the shell should be able to do file completion without issues.
Depending on how you have zsh configured, it may also work to use setopt MAGIC_EQUAL_SUBST. I normally have that option set and get the behavior that you are looking for, but get the behavior that you currently experience if I unset that option. But, if I use zsh -f to skip loading my configuration just enabling that option isn't sufficient so there is apparently something else in my config that was necessary, possibly just the enabling of the completion system which you would likely have done already. There could be cases where this would result in undesirable behavior, but I haven't run into any in the many years that I've been using zsh with that option set.
A more complicated option would be to write a completion function for that command which tells zsh that that option takes a file name as the argument.
I am implementing my own shell. But to supprt for command history, I need to use monitor up/down arrow key as in standard shells. Please tell me how to handle arrow keys as input or does these keys generate signals? Please elaborate.
Arrows and other special keys send you special strings that depend on the terminal being used or emulated. To deal with this most simply, you could use a library such as termcap. Even simpler, given your stated purpose (command history support), would be to use readline, which basically does it for you (and lets the user customize aspects of their preferred mode of working across the many applications that link to the same library).
It depends on how gnarly you're expected to go. Worse case, you're writing the keyboard interrupt handler. Best case, something like readline.
Check with your prof for direction here. Also check your course materials to see if the prof has given links/examples regarding this.
Does the assignment specifically say you need to have a "cursor key driven" command history?
Simple option is to mimic the shells fc e.g.
$ ls
... file listing ...
$ fc -l
1 ls
2 fc -l
$ fc -r 1
... file listing ...
and (while I'm presenting established ideas as my own, might as well go all the way) to be able to edit the command line you could use
fc -e start end
to write the history from start to end into a file, launch an editor and then execute the resulting file as a script. This way your shell is not using a library, but launching commands and executing scripts, which is what shells are supposed to do anyway.
Do you know any resources that teach to good habits of working in UNIX command line?
EDIT: I don't mean general books about shell or man pages. I mean the things that you can only see watching professionals working with command line. For example when changing frequently between two directories they use "pushd" command, when repeating a command they use "history". I can read about these commands but I want to make it a habit to use them effectively.
I am speaking out of my own experience so it may not apply to you;
The best way to be efficient is actually using it on a daily basis, instead of using graphical tools even if they make look things easy. You will then become aware of most common tasks you care about, and instead of trying to grok it at once, you get a fairly good starting point to start learning. Man pages are the first thing to look at, but there will be non-obvious tricks which you need to search anyway. Knowing what you exactly want, infinitely increases probability of finding it.
For example, you can find how to search all mp3 files easier in man page of "find" than how to deal with files in general (where to start?).
Some common bash command line actions, not in order:
Command line editing: you'll want to be good with emacs or vi and apply that to editing your commands.
Completion: use TAB to expand file names and paths.
note: There is a huge set of file, command, and history completion functions, and it is configurable. Big topic.
"cd -" : go back to the last directory you were in
~ = home directory (or ~user for users home dir)
"ESC ." : expands to the final arg from the previous command
"!string" : execute the last command starting with string
learn find, grep, sed, piping "|" and redirection ">". You'll often combine these to do useful things.
Loops from the shell prompt, e.g. "for" loop - to do repetitive actions
Learn your regular expressions! Often used for matching files.
example: ls x[0-5]*.{zip,tar} = list files starting with x, followed by a number 0 through 5, followed by any string ending in .zip or .tar
If possible ask others for their favorite tricks, read the manual, and practice.
For the more advanced stuff This seems to be fairly comprehensive
this is a great resource: "Rute User's Tutorial and Exposition" (http://rute.2038bug.com/index.html.gz)
stackoverflow.com esp. the bash tag ;-)
(and of course the bash man page)
If you want things that you can "only see watching professionals working with command line," then you've answered your own question: Watch professionals working with the command line. I don't personally find that very useful unless the other person is doing the same thing multiple times; it's hard to pick something up after just one session because it's hard to watch the screen and the keyboard at the same time.
I think the key is to not try to become an expert right away. Just use the command line frequently, and be aware that you might not be using it as well as you could, but don't let that discourage you from using it anyway.
Browse through the man page of your shell, and through lists of tips, not with the goal of memorizing everything in them, but just to pick out a couple of things to try out. Skim through until something catches your eye and makes you think, "Gee, that sounds useful." Then try it out. Not everything is going to be useful immediately; you might have to wait a while before you encounter a situation where you can try something out. Maybe you could write down some things on Post-It notes by your desk to remind you that certain feats are possible, so when you encounter a situation where a more obscure feature could be handy, you'll be more likely to remember to try it.
Frankly, it's impossible to learn this stuff in a vacuum. You need to have problems to solve.
While it certainly helps to have familiarity with the tools available (of which there are a myriad), "learning" it requires applying it. And applying it requires "real" problems to solve.
For example, the skillset of a System Admin may be different from someone who works with databases because their roles are different.
I use them for data processing, using mostly one off files. /tmp/x.sh and /tmp/x.x are worn bare in the directory folder.
My hammers tend to lean towards: ls, find, sort, sed, vi, awk, grep, and comm. Combined with simple shell scripting like: for i in cat /tmp/list; do .. done
But I do a lot of ETL work, and very few script files, which is why my shell scripting skills are so weak.
I do rely on one script, however:
#!/bin/sh
# latest -- show latest files
ls -lt $# | head
As 95% of the time the files I'm working on are in the top 10 latest files. And "latest *.txt" works a peach.
So, bottom line, you need problems to solve. You need to learn the 'man' command, man -k is nice to find things. You also need to leverage the "See Also" at the bottom of most man pages. That's a treasure trove of "I didn't know you could do that".
Then, just start solving problems. Start figuring out "what would be nice to have" and then see if it exists (it very well may). If not, awk, perl, or python can make those "nice to haves" out of thin air.
Join a LUG. That is where I learned most things early on. Ask the organizers to do a "Bash Tips And Tricks Night".
Deft shell users love to show off.
apropos is a really good tool for this sort of thing. Whenever you find yourself unsure of the best way to do something, or wishing you weren't repeating yourself, just use apropos with a keyword or two to find other commands that can help. In distros like debian, you can also install web-based help tools that search all of the manuals available on the system: texinfo, man pages, html, and pdf etc.
Aside from that, yep, read your shell's manual right through at least once --- preferably, go back to repeatedly it as you learn more, reach limits and want to be more efficient.
The join a LUG idea is also good; you'll definitely learn from others' demos.