Tmux

Terminal Multiplexer

• for “serious work” on a remote server
• several virtual windows in one shell
• switching between “windows”
• persistent sessions (all windows)
  • lost connection can be re-started later
  • all windows and their hisrories are restored
• resume session from a different machine
  • continue exactly where you left off
• replacement of the older tool screen
  • tmux is more capable and easier to use

    tmux                   # start tmux
        ls                 # do something
        <C-b> %            # split the window
        less /etc/passwd   # do sth in the split
        <C-b> d            # detach from tmux (keeps running)
    ...
    tmux attach            # attach to the running tmux

Tmux (continued¹)

vertical window split

Tmux (continued²)

horizontal window split

Tmux (continued³)

prefix indicates a command for Tmux:

Text editors

• MS Word is not a text editor (it is a word processor)
• Editing plain text (storing no formating like sizes, colors, etc. in the document/file
• Mainly for programming
• Good for any text processing needing exact text control
• Anything can be opened as raw text/binary by text editors
• Good editors, but only working in GUI (need to know a backup editor for ssh): - Atom, Visual Studio Code (VS Code)
• Most well known and capable (work in terminal too): vim, emacs
  • both have steep learning curve, but once learned, they are very efficient
• joe: good compromise: quite capable and quite easy to use
• pico, nano: minimalistic editors. Easy to use as emergency in terminal, but not suitable for serious work
• Pick one (universal) or two (GUI for local use + something for terminal) and learn to use them.

Setting up your Linux on Windows

• If you are using WSL and a linux on Windows (like Ubuntu or Suse), you can make these comfortable shortcuts:
• ln -s /mnt/c/… (path to your linux home) winhome (more about links later)
  • now you can do for instance cd winhome
• add useful shortcuts to your ~/.bashrc
  • you can always use notepad.exe ~/.bashrc, but there really are better editors 😀
    • alias Less=less -M
    • alias ll='ls -lh'
• You should also decide:
  • Do you want to access your files also from Windows and with Windows applications? If so, do all your work in the Windows drive, not the linux home, which is hidden from Windows
    
• You can use a native Windows text editor in this case
• Atom, or Visual Studio Code are available as Windows applications
• code . opens the current directory in VS Code
• Working across linux / windows borders brings speed penalty. Usually not important, but keep it in mind. For demanding work, move data to your linux home before processing via linux tools.

Any Good Text Editor™

Modern source code editor should provide:

• modes (programming languages, xml, html…)
• syntax highlighting
• completion
• indentation
• support for encodings (utf-8, but also older)
• undo (multiple)
• searching (with regexp)
• integration with shell, compiler and/or debugger
• fallback mode for working in a text console (no GUI)

(source: NPFL092)

Vim – Vi improved

• accessed by default from within less by command v
  • (or any other editor set in $EDITOR)
  • if started by accident, get out by typing <esc>:q!
• modal editor … user modes
  • normal mode (not so normal) <esc>
  • input mode (rather normal) i
  • visual mode v/V
  • and several more
• vimtutor
• high quality help system :h help
• An older but good free book in PDF
  • written for Vim 5.7 in 2001, so consult documentation for new features like omni-completion or spell-checking

vim +nn <file>

The cursor starts on the specified line.

Program

• Sequence of commands
• conveniently saved into files
• chmod +x on a file makes it “executable”, so it can be run as ./my-program.pl
• semicolon (;) separates commands
• special first line tells shell what to execute
  • #!/usr/bin/env bash
  • #!/usr/bin/env bash

#!/usr/bin/env bash
# a silly program (script) that generates another small program
echo '#!/usr/bin/env bash' >My_Sort.bash
echo 'sort | uniq -c | sort -nr' >>My_Sort.bash
chmod +x My_Sort.bash # now it is executable!

last | cut -f1 -d' ' | ./My_Sort.bash

POS tagging

• Universal Dependencies: unified representation for dependency treebanks
  • both syntactic (how it is written), and semantic (what it means) unification
  • http://universaldependencies.org
• Treebanks in the CoNLL-U format
  • simple plain text (UTF-8) tables
  • sentences separated by empty lines
  • # comment lines;
    • some comments are special, like plaintext representation of the sentence
  • one word per line
  • \t separates columns
  • columns with “node” attributes:

word order, form, lemma, POS, morphology-features, parent, ...

POS tagging Ⅱ

• take Czech training data from UD 2 Czech
• get form, lemma, POS
• Write your own rule-based POS-tagger
  • Rules should generate (and rewrite) 5th column with POS
• Write an evaluation tool to compare 4th and 5th column
  • Use the tool to evaluate an impact of each new rule
  1. formulate hypothesis (rule)
  2. Evaluate / test hypothesis
  3. Accept or reject