I have recently been faced with a rather odd task, one result being the necessity for the ability to use DTMF (aka "Touch Tone") tones to control a non-X Linux computer's terminal. The computer has a modem which can be accessed through ALSA, and therefore the sox "rec" program, which is what I am reading the input through. The computer in question is otherwise completely isolated, having no Ethernet or other network interfaces whatsoever. The Goertzel algorithm implementation I am using works very well, as does the eSpeak speech synthesis engine which is the only source of output; this is supposed to work with any Touch Tone phone. It reads back both input (input being octal digits, one ASCII byte at a time)and whatever the dash shell feeds back -- the prompt, the output from commands, etc., using ASCII mnemonics for control characters.
The current method that I am using for interacting with dash and the programs launched through it is the pexpect module. However, I need it to be able to, on demand, read back the entire contents of the line on which the cursor is positioned, and I do not recall pexpect being able to do this (If it is, I cannot tell.). The only other solution that I can think of is to somehow use Python to either control, or act as, the keyboard and console drivers.
Is this, indeed, the only way to go about it (and if so, is it even possible with Python?), or is there another way of having direct access to the contents of the console?
Edit: Through dumb luck, I just recently found that the SVN version of PExpect has pexpect.screen. However, it does not have any way of actually running a program under it. I'll have to keep an eye on its development.
The simple solution is to use the Linux kernel uinput interface. It allows you to insert keypresses and mouse events into the kernel, exactly as if they came from a physical human interface device. This would basically turn your application into a keyboard/mouse.
Since you are working with Python, I recommend you take a look at the python-uinput module.
If you are comfortable with binary I/O in Python, then you can certainly do the same without any libraries; just check out the /usr/include/linux/uinput.h header file for the structures involved (the interface is completely stable), and perhaps some uinput tutorials in C, too.
Note that accessing the /dev/uinput or /dev/input/uinput device (depending on your distribution) normally requires root privileges. I would personally run the Python service as a user and group dedicated to the service, and modify/add a udev rule (check all files under rules.d) to allow read-write access to the uinput device to that group, something like
SUBSYSTEM=="input", ENV{ID_INPUT}=="", IMPORT{builtin}="input_id"
KERNEL=="uinput", MODE="0660", GROUP="the-dedicated-group"
However, if your Python application simply executes programs, you should make it a terminal emulator -- for example, using this. You can do it too without any extra libraries, using the Python pty; the main work is to however simulate a terminal with ANSI escape sequences, so that applications don't get confused, and the existing terminal emulators have such code.
If you want to manipulate the contents of the console, you probably want to use curses. It's well documented here. Look at window.getch() and window.getyx().
Related
I just made a little chat CLI-application that works pretty fine actually!
Unfortunately, when one types a message and receives other's messages at the same time, both interlacing... Which renders quite awful:
elton: Hey!
john: how are you doing?
fine anjohn: still not to bed?!
d john: haha
you?elton: fine and you?
I'm looking for a way to avoid this kind of problem. Such as "reserving" the last line for user-input or process some actions when data are received to recalculate the position of user-input.
After some research I found that I shall retrieve each character one by one with getch(). So that I can check regularly if a new message is waiting to be displayed and handle this case.
But if I use getch() I have to redefine manually basic actions (such as backspace, move left and right...), some characters take more than one byte. In brief, it's unusable.
As far as I can tell, you have a few options here. Normally just calling getch you're running the terminal in what's called "cooked" mode. This means that the terminal just displays characters in the order they arrive, nothing too special. However you run in to race conditions as you've discovered.
Option 1: you read things in one character at a time, buffering the entire line. When a new line arrives and you want to print it out, you'd have to (a) grab hold of some kind of print mutex (internal to the program), (b) clear out the current line (print '\r[space][space][space]...\r'), (c) print the incoming line + '\n', and (d) restore the previous buffer to the screen (via print) and unlock the mutex.
This gets ugly. Fast. As you discovered, there's no line editing support or anything fancy. The only good thing about this approach is that it'll probably work in about 99% of terminals.
Option 2: you put the terminal in to raw mode. In this mode, you have complete control over the terminal, but you lose very basic functionality (e.g. typing a key won't display it unless you manually output it). You would still have to manually implement things like navigation, but at least it would work. This method is probably the most flexible and give you the most control, however it's very difficult to work like this, which is why they invented...
Option 3: you use ncurses (or a similar terminal library). The concepts are a little difficult, but FAR easier to learn than doing things in pure raw mode. Editing and windowing (or as you put it, reserving a line) are built-in. The interface can be much prettier than you'd get from cooked mode.
EDIT I'm talking a lot about *nix terminals here, most of which doesn't apply for Windows.
The usual way to do this kind of thing is by writing a "console-mode GUI" instead of a pure CLI.
For example, you can very easily create a very simple interface with a ROWS-2xCOLS output window, a 1xCOLS divider or mode line, and a 1xCOLS input window—as seen in things like IRC clients, emacs, etc. For example:
Or you could do a split view, with a ROWS/2xCOLS window for the other person/people's chat, and a ROWS/2xCOLS window below that for the user's current and previous messages, as seen in most of the early Unix chat programs.
Or something fancy with windows all over the place that pop up and go away as appropriate, as some modern IRC clients do.
Python has curses support built-in—but not on Windows. If you want a cross-platform solution, or a Windows-only solution, see curses-like library from cross-platform console app in python, and the answer is probably there.
Here's a screenshot from the IRC client BitchX, to give you an idea of what's possible without too much work:
I would like to launch an untrusted application programmatically, so I want to remove the program's ability to access files, network, etc. Essentially, I want to restrict it so its only interface to the rest of the computer is stdin and stdout.
Can I do that? Preferably in a cross-platform way but I sort of expect to have to do it differently for each OS. I'm using Python, but I'm willing to write this part in a lower level or more platform integrated language if necessary.
The reason I need to do this is to write a distributed computing infrastructure. It needs to download a program, execute it, piping data to stdin, and returning data that it receives on stdout to the central server. But since the program it downloads is untrusted, I want to restrict it to only using stdin and stdout.
The short answer is no.
The long answer is not really. Consider a C program, in which the program opens a log file by grabbing the next available file descriptor. Your program, in order to stop this, would need to somehow monitor this, and block it. Depending on the robustness of the untrusted program, this could cause a fatal crash, or inhibit harmless functionality. There are many other similar issues to this one that make what you are trying to do hard.
I would recommend looking into sandboxing solutions already available. In particular, a virtual machine can be very useful for testing out untrusted code. If you can't find anything that meets your needs, your best bet is to probably deal with this at the kernel level, or with something a bit closer to the hardware such as C.
Yes, you can do this. You can run an inferior process through ptrace (essentially you act as a debugger) and you hook on system calls and determine whether they should be allowed or not.
codepad.org does this for instance, see: about codepad. It uses the geordi supervisor to execute the untrusted code.
You can run untrusted apps in chroot and block them from using network with an iptables rule (for example, owner --uid-owner match)
But really, virtual machine is more reliable and on modern hardware performance impact is negligible.
I like the python-send-buffer command, however I very often use Python embedded in applications, or launch Python via a custom package management system (to launch Python with certain dependencies).. In other words, I can't just run "python" and get a useful Python instance (something that python-send-buffer relies on)
What I would like to achieve is:
in any Python interpreter (or application that allows you to evaluate Python code), import a magic_emacs_python_server.py module (appending to sys.path as necessary)
In emacs, run magic-emacs-python-send-buffer
This would evaluate the buffer in the remote Python instance.
Seems like it should be pretty simple - the Python module listens on a socket, in a thread. It evaluates in the main thread, and returns the repr() of the result (or maybe captures the stdout/stderr, or maybe both). The emacs module would just send text to the socket, waits for a string in response, and displays it in a buffer.
Sounds so simple something like this must exist already... IPython has ipy_vimserver, but this is the wrong way around. There is also swank, while it seems very Lisp-specific, there is a Javascript backend which looks very like what I want... but searching finds almost nothing, other than some vague (possibly true) claims that SLIME doesn't work nicely with non-Lisp languages
In short:
Does a project exist to send code from an emacs buffer to an existing Python process?
If not, how would you recommend I write such a thing (not being very familiar with elisp) - SWANK? IPython's server code? Simple TCP server from scratch?
comint provides most of the infrastructure for stuff like this. There's a bunch of good examples, like this or this
It allows you to run a command, provides things comint-send-string to easily implement send-region type commands.
dbr/remoterepl on Github is a crude proof-of-concept of what I described in the question.
It lacks any kind of polish, but it mostly works - you import the replify.py module in the target interpreter, then evaluate the emacs-remote-repl.el after fixing the stupid hardcoded path to client.py
Doesn't shell-command give you what you are looking for? You could write a wrapper script or adjust the #! and sys.path appropriately.
I have a Panatone Huey, a monitor calibration probe (device you attach to the monitor, and it gives you colour readings) - I want to get readings from the device in Python.
Having never written such a device driver before, I'm not sure where to start.
I've found are two open-source C/C++ projects that interface with the Heuy - ArgyllCMS and mcalib.
ArgyllCMS comes with a spotread command which returns readings from the device, although it only functions as an interactive command line tool, so running it via subprocess will not (easily) work.
The code ArgyllCMS uses to communicate with the device is in spectro/huey.c
Not tried it (only just found it while writing this question), but mcalib contains much less code, mainly just heuy.cpp - however it has a worrying number of FIXME comments and incomplete methods, and the code appears to have been automatically generated (unhelpful variable names)
There seems to be three options:
Modify spotread to work without any interactive prompts, call it via subprocess
Create a C-based Python module around huey.c or huey.cpp
Re-implement the interface using something like PyUSB
Being much more familiar with Python, I'm tempted to use PyUSB, but will this be substantially more work than wrapping existing code with the Python C API? Is there anything obvious in either of the C implementations that will not be easily doable in PyUSB?
Given the existence of spotread the easiest (though perhaps not the best) way to proceed would be to use pexpect. It allows you to interact with other command-line programs.
I am trying to write a cross-platform python program that would run in the background, monitor all keyboard events and when it sees some specific shortcuts, it generates one or more keyboard events of its own. For example, this could be handy to have Ctrl-# mapped to "my.email#address", so that every time some program asks me for my email address I just need to type Ctrl-#.
I know such programs already exist, and I am reinventing the wheel... but my goal is just to learn more about low-level keyboard APIs. Moreover, the answer to this question might be useful to other programmers, for example if they want to startup an SSH connection which requires a password, without using pexpect.
Thanks for your help.
Note: there is a similar question but it is limited to the Windows platform, and does not require python. I am looking for a cross-platform python api. There are also other questions related to keyboard events, but apparently they are not interested in system-wide keyboard events, just application-specific keyboard shortcuts.
Edit: I should probably add a disclaimer here: I do not want to write a keylogger. If I needed a keylogger, I could download one off the web a anyway. ;-)
There is no such API. My solution was to write a helper module which would use a different helper depending on the value of os.name.
On Windows, use the Win32 extensions.
On Linux, things are a bit more complex since real OSes protect their users against keyloggers[*]. So here, you will need a root process which watches one of[] the handles in /dev/input/. Your best bet is probably looking for an entry below /dev/input/by-path/ which contains the strings "kbd" or "keyboard". That should work in most cases.
[*]: Jeez, not even my virus/trojan scanner will complain when I start a Python program which hooks into the keyboard events...
As the guy that wrote the original pykeylogger linux port, I can say there isn't really a cross platform one. Essentially I rewrote the pyhook API for keyboard events to capture from the xserver itself, using the record extension. Of course, this assumes the record extension is there, loaded into the x server.
From there, it's essentially just detecting if you're on windows, or linux, and then loading the correct module for the OS. Everything else should be identical.
Take a look at the pykeylogger source, in pyxhook.py for the class and implimentation. Otherwise, just load that module, or pyhook instead, depending on OS.
I've made a few tests on Ubuntu 9.10. pykeylogger doesn't seems to be working. I've tryied to change the /etc/X11/xorg.conf in order to allow module to be loaded but in that specific version of ubuntu there is no xorg.conf. So, in my opiniion pykelogger is NOT working on ubuntu 9.10 !!
Cross-platform UI libraries such as Tkinter or wxPython have API for keyboard events. Using these you could map «CTRL» + «#» to an action.
On linux, you might want to have a look at pykeylogger. For some strange reason, reading from /dev/input/.... doesn't always work when X is running. For example it doesn't work on ubuntu 8.10. Pykeylogger uses xlib, which works exactly when the other way doesn't. I'm still looking into this, so if you find a simpler way of doing this, please tell me.
Under Linux it's possible to do this quite easily with Xlib. See this page for details:
http://www.larsen-b.com/Article/184.html