I've seen strange readline editing behavior when the editing doesn't start at the first column: I can actually backspace into the prepended message (with Del, Ctrl-W or Ctrl-U), and then the whole edit becomes messed up.
read can output a prompt on its own (hopefully in all versions of Bash that we aim to support - the tests will tell), and that doesn't have this problem, and it's also a bit cleaner and shorter.
The prompt is only displayed if input is coming from a terminal. For the tests (currently only deletion and move confirmations are covered), this means that the prompt itself cannot be covered, and an empty line instead has to be expected. (On the positive side, this removes the ugly trick with $SPACE.)
If a custom action cannot be made non-executable, it needs to be removed as well (and the test skipped); otherwise its existence will break following tests that assume it's inactive.
So that any combination of priority / date entered in the replacement will replace the corresponding original ones, but if they are left out, the original ones will be kept.
In essence, omitted stuff will be kept, added stuff will override, only deletion of existing stuff is not possible (but this is replace, after all).
Fixes#386
Instead of potentially falling back to the built-in action that a custom action was intended to override, but (e.g. due to file system reorganizations) now results in a broken link. The extension functionality that is then skipped may result in undesired results, but this may not be immedately obvious to the user (if the extension is not particularly verbose), so some data corruption could occur if this remains undetected.
To avoid duplicating (or somehow extracting) all the built-in actions, simply detect _any_ broken symlink; i.e. offer a superset of the required functionality. So this would also complain about a broken symlink to a non-executable custom (auxiliary) file (rarely used) if that is mistakenly passed as a custom action (unlikely).
Fixes#359
* fix whitespace
* fix spelling
* fix whitespace
* unify headers of tests
* fix some issues in tests, identified by shellcheck
* fix bash completions
bash completion files are not supposed to be executable
* fix some issues identified by shellcheck
Co-authored-by: Ali Karbassi <ali@karbassi.com>
* Tests: Add coverage for del / move without -f, but with prompting
Supplying the user confirmation via "yes".
* Cosmetics: Align inconsistent spacing for before (y/n) prompt
* Refactoring: Extract confirm() function
The user confirmation query had been duplicated (once) in the code.
* Refactoring: confirm(): Leave early if forced
* Return from user prompt without requiring Enter
By just reading a single character (y for yes, anything else: no).
* Tests: Ensure that only a single "y" concludes the confirmation
By switching from "yes" (that endlessly prints newline-separated "y"s) to "printf y".
* t1800-del: Add coverage for negative confirmation
Negative means "anything but y", so "n", "x", and Enter all apply.
* Cosmetics: Add trailing space after (y/n) prompt
So that the user's typed answer is not recorded directly after it, but with separation: "Foo? (y/n) y" instead of "Foo? (y/n)y".
*Compatibility: "read -N 1" is only available in Bash 4.1+
Mac OS still ships with Bash 3.2 :-( Fall back to the original prompting that requires conclusion via Enter then.
Note: Even though the tests use "printf y", this still gets accepted, as there'll be EOF after that. In real use (when stdin from the terminal stays open), a concluding Enter is mandatory, though.
Closes#152
If the current directory had spaces in the directory path, this script would fail.
That has been fixed by appropriate use of quotes.
Co-authored-by: Tim Learmont <learmont@NocturnalAviation.com>
Co-authored-by: Ali Karbassi <ali@karbassi.com>
Uses current priority logic and option tests just running multiple times until item and priority pairs are exhausted or stops on first error.
Tests for multiple priorities, multiple re-prioritization and a couple of errors.
* TESTSFIX: t0000-config.sh may unintentionally pick up existing user config in `$XDG_CONFIG_HOME/todo/config`
This corner case was missed when the XDG alternative config location was introduced, likely because the variable isn't set by default on Ubuntu or Mac OS, and none of the developers had a config there.
HOME is already redirected to the current working directory; as `XDG_CONFIG_HOME` is just a config alternative that has a fallback to `$HOME/.config`, simply unset it.
Fixes#347
* TESTSFIX: User add-ons that override built-in actions may break tests
If `TODO_ACTIONS_DIR` is already defined, the defaulting in todo.sh are skipped. The user may have custom add-ons that override built-in actions; any incompatible behavior would negatively affect the tests then.
By setting the variable to the first default location, we avoid this interference.
* ENH: Allow to customize what can be before / inside / after +project and @context
I noticed that todo.sh mistakenly picks up some random stuff as projects and contexts. The todo.txt specification is very lenient in what makes those (allowing any length of non-whitespace characters), but in practice, every user has its own idea of the set of used words. This change allows (expert) users to tweak the pattern (via TODOTXT_SIGIL_VALID_PATTERN), to be able to reduce the amount of "false positives" that listproj and listcon show.
Additionally, there's a (mainly unspecified) extension mechanism in the form of key:value, which often is used to implement custom "markers", like w:something to make a task blocked and waiting for "something". It would be nice to be able to refer to @contexts (and maybe also +projects), but the default syntax (which requires whitespace around those) doesn't allow it, and these context then don't show up in searches, which is unfortunate. The added TODOTXT_SIGIL_BEFORE_PATTERN and TODOTXT_SIGIL_AFTER_PATTERN allow to tweak the syntax to accommodate this.
The implementation just exchanges grep with a corresponding sed pipeline step, which shouldn't negatively affect the performance at all (as sed is one of the main commands used thoughout the script), but which allows for the additional matching flexibility.
* FIX: Compatibility: sed \? multi not supported on MacOS
Use the POSIX \{0,1\} instead.
* FIX: Compatibility: sed \| multi not supported on MacOS
Concatenate both alternatives as optional matches; this is only an approximation but satisfies the test case here. It would probably be better if the entire sed code used extended regular expression syntax (-E) instead of basic ones (-e), but that change would also affect all add-ons. As this is an optional customization feature, uses who really need this either also have to work around, or supply a GNU sed port on MacOS.
* FIX: Compatibility: sed \+ multi not supported on MacOS
Use the POSIX \{1,\} instead.
Co-authored-by: Ali Karbassi <ali@karbassi.com>
* Refactoring: Use regular expression match instead of case globbing for actions taking a SRC argument
Making it consistent with the test for MOVE_COMMAND_PATTERN, and allowing to extend the pattern with custom actions in the future.
* Refactoring: Move the anchoring and grouping out of MOVE_COMMAND_PATTERN
So that additional (custom) add-on actions can be configured in the future.
* ENH: Enable file completion for add-on actions via _todo_file{1,2,3}_actions
Allowing completion of todo files directly after the add-on action (1), with one (2) / two (3) arguments in between. This should handle most cases.
In order to configure the add-on completion, the corresponding configuration variable has to be defined in the user's shell (e.g. via ~/.bashrc):
_todo_file1_actions='myaction|anotheraction'
* add highlighting of dates, item numbers, and metadata
* add test for highlighting of dates, item numbers, and metadata
Co-authored-by: Ali Karbassi <ali@karbassi.com>
* Handle missing EOL in todo.txt
This can happen easily with certain editors (such as Mousepad) that do not automatically add a newline character at the end of a file.
In _addto(), ensure a trailing newline via sed (taken from https://unix.stackexchange.com/a/31955/18876).
Fixes#294
* Tests: Add basic coverage of move
* Handle missing EOL in todo.txt for move, too
This can happen easily with certain editors (such as Mousepad) that do not automatically add a newline character at the end of a file.
* Refactoring: Extract fixMissingEndOfLine()
* FIX: Compatibility: sed \+ multi not supported on MacOS
Use the POSIX \{1,\} instead.
Co-authored-by: Ali Karbassi <ali@karbassi.com>
- In macOS 10.15, the `date` command adds a "-R" flag. This can be seen in the default error display. Since the flags have changed, the grep command to find which version of macOS date binary we have. Because of this change, we need to do our own check for macOS 10.15.
- Remove windows from the github actions as this tool is not for Windows, currently.
Adds the config / env variable `TODOTXT_PRIORITY_ON_ADD`, which when set to one of the capital letters `A` to `Z` is given as a priority to a new task, if a priority is not given on command line.
Addons can be placed in subfolders with the same name as the action,
in order to be able to e.g. clone git repos into the TODO_ACTIONS_DIR
rather than having to download addons and manage updates manually.
Closes#120
Because @contexts and +projects are metadata as much as they are text,
it can be nice to highlight them independently of the coloring of the line
that contains them. This is tough to do correctly in
a TODOTXT_FINAL_FILTER because doing it at that layer requires
re-parsing the text of the line to find the appropriate PRI_$x color to
reset to after coloring the word, or trying to reverse engineer where
the color codes inserted by todo.sh end and the user's text begins (and
vice versa).
Closes#119
Request from exatto (http://article.gmane.org/gmane.comp.diary.todotxt/4699), who doesn't want to use done.txt and report.txt.
Also consider character devices (-c; /dev/null is one) in the file existence checks so that no empty file is then created.
If not found elsewhere, look for the config file in /etc/todo/config.
`make install` installs a global config file, setting TODO_DIR=~/.todo.
todo.sh creates TODO_DIR if it doesn't exist.
The user may need the help to solve any fatal error that appears while todo.sh isn't properly set up. As the help actions do not depend on any setting that the fatal errors check, we can still invoke them.
Factor out dieWithHelp() and use that for printing the fatal errors.
Need to make match of preceding priority optional.
Also tighten patterns for priority and done marker by anchoring them to the front, as the leading task number has already been substituted away.
When the command-under-test fails with a non-zero exit code, neither the different exit code nor any output differences are printed. Also, there's a lot of duplication between test_expect_output and test_expect_code_and_output. So, let's fix both by delegating the first to the latter.
Currently, the only way to get usage help for a custom action is to use "todo.sh help" and scroll / search for the action name. (Or try to call the action without / with invalid parameters to hopefully get a one-line syntax summary.)
This extends the help command to print the usage of optionally passed custom action names.
This applies e24777fa2c to todo.cfg, too, and adds a verification to the existing test.
Note: Because todo.cfg is preprocessed by test-lib, we use the original config from the working copy, not the copy inside the test fixture.
An empty line is used to delimit test commands in test_todo_session. (This wasn't properly reflected in the usage comment, fixed that, too.) This prevents us from testing commands that include empty lines. (Currently, there is no such output in todo.sh, but custom add-ons may have this.)
Instead of forcing the test to filter out the empty lines, extend the syntax to allow empty lines in the expected output by escaping them with a single backslash. (When redirecting via <<EOF, the backslash itself must be escaped = doubled; better use <<'EOF'.)
