Implement terminal_pager for log subcommand #46
Conversation
|
|
||
|
|
||
| bool m_grabbed; | ||
| std::ostringstream m_oss; |
There was a problem hiding this comment.
It seems that m_oss is used only to get its internal buffer. Why not directly store a std::stringbuf instead?
| #include "terminal_pager.hpp" | ||
|
|
||
| terminal_pager::terminal_pager() | ||
| : m_grabbed(false), m_rows(0), m_columns(0), m_start_row_index(0) |
There was a problem hiding this comment.
std::cout.rdbuf(std::cout.rdbuf()) is a noop, so I think we could always capture std::cout buffer, and always restore it in release_cout, even if it was not redirected. This way we could drop m_grabbed and the condition in the implementaion of release_cout.
| // Unfortunately need to access _internal namespace of termcolor to check if a tty. | ||
| if (!m_grabbed && termcolor::_internal::is_atty(std::cout)) | ||
| { | ||
| // Should we do anything with cerr? |
There was a problem hiding this comment.
We should probably capture std::cerr too as by default (at least on some platforms) it outputs to the same place as std::cout
| void maybe_grab_cout(); | ||
|
|
||
| // Return true if should stop pager. | ||
| bool process_input(const std::string& input); |
There was a problem hiding this comment.
I think it's fine to take the argument by value; calling it with an rvalue will trigger the string's move constructor.
| auto end_row_index = m_start_row_index + m_rows - 1; | ||
|
|
||
| std::cout << "\e[2J"; // Erase screen. | ||
| std::cout << "\e[H"; // Cursor to top. |
There was a problem hiding this comment.
Should we capture these "magic strings" into constants with expressive names?
| return false; | ||
| case '\e': // ANSI escape sequence. | ||
| // Cannot switch on a std::string. | ||
| if (input == "\e[A" || input == "\e[1A]") // Up arrow. |
| auto new_termios = old_termios; | ||
| // Disable canonical mode (buffered I/O) and echo from stdin to stdout. | ||
| new_termios.c_lflag &= (~ICANON & ~ECHO); | ||
| tcsetattr(fileno(stdin), TCSANOW, &new_termios); |
There was a problem hiding this comment.
I think this and the call to tcsetattr to restore the initial state line 211 should probably be in a scope object to guarantee the initial state is restored even if an excpetion is thrown.
The start of a terminal pager for various
git2cppcommands, initially enabled forlogsubcommand. Fixes #45.pager.mp4
Supports the following keys to navigate:
If cout is not to a tty or the output is short enough to fit within a single page the pager is not used.
Still to do, probably in separate PRs:
To add to other subcommands, see how it is used in
log_subcommand.cpp. Theterminal_pagerconstructor does the required initialisation, and it displays on theshowcall. I suppose I could have done theshowautomatically in the destructor but that didn't seem a good idea.Awaiting a new release of JupyterLite before we can easily try it in the terminal.