One in the fuzz check handling of NaN's and a pathway to generate a
HorribleError.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Fixes#1139
Fix a bug relating to parsing negative floating point values with no
zero before decimal point in a context where an option could be
possible.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
max, min on positional was not respected, specifically max positionals
only filled up to min, and skipped over optional options.
Fixes#1090
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Move the help generation priority higher so it triggers before config
file processing.
Fixes#1099
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This has been bounced around for a couple years now
#474 and a few others have expressed desire to work with non-standard
option names. We have been somewhat resistant to that but I think it can
be done now. This PR adds a modifier `allow_non_standard_option_names()`
It is purposely long, it is purposely off by default. But what it does
is allow option names with a single `-` to act like a short option name.
With this modifier enabled no single letter short option names are
allowed to start with the same letter as a non-standard names. For
example `-s` and `-single` would not be allowed.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Henry Schreiner <HenrySchreinerIII@gmail.com>
Add modifier for subcommands to restrict subcommands falling through to
parent.
This will resolve#1022
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
_This is the new PR I've mentioned to work on in PR #858_
## A better Help Formatter
_See below for images of the new help page_
Finally, after a lot of planning, understanding CLI11's codebase,
testing and coding, the new default Help Formatter is done. There are a
lot of changes to make the help page more readable and closer to UNIX
standards, see Changelog below for details. One of the highlights is
automatic paragraph formatting with correct line wrapping for App and
options/flag descriptions as well as the footer.
A goal was to provide more flexibility and better readability for the
help page while providing full compatibility with Apps using CLI11 (no
breaking changes and no changes to Apps required). Also better support
for different terminal sizes. Users can now specify three new optional
attributes: `right_column_width_`, `description_paragraph_width_` and
`footer_paragraph_width_`. See code documentation for more details. The
different columns for options/flags now scale with the set
`column_width_` value: Single dash flags occupy 33% of the set
`column_width_`, double dash flags and options (like REQUIRED) 66%.
These new attributes allow for indirectly respecting terminal geometry,
footer paragraph formatting has also been added (#355). This PR also
implements the issues #353 and #856.
The new help page formatting can also be used as an input for man page
generation, since it's oriented on the man page style (#413).
[help2man](https://www.gnu.org/software/help2man/) can be used to
generate man pages from help output (see comment down below for
example).
I thoroughly tested this code with all possible combinations of flags,
options, positionals, subcommands, validators, ...
So far everything works great.
I hope this PR looks good and meets all requirements. I'm looking
forward to the implementation of this PR into CLI11. If you have any
questions or suggestions feel free to comment.
### Fixed/implemented issues by this PR
- #353 Better options formatting
- #856 Space between options
- #355 Footer formatting
- #413 Man page generation can be achieved using help2man with the new
help formatting
- https://github.com/CLIUtils/CLI11/issues/384#issuecomment-570066436
Better help formatting can be marked as complete
### What about the failing tests?
Of course the tests expect the old help text format. This is why 6 of
the tests are failing. Since it is a bit of work to migrate the tests to
the new help format, I first wanted to push out this PR and get
confirmation before I'll update all the tests.
So please let me know if this PR gets implemented, what changes should
be made and then I'll migrate the tests to the new help format, either
in this PR or I'll make a new one.
## Changelog:
#### There are _no breaking changes_. Every App using CLI11 will work
with this new formatter with no changes required.
- Added empty lines at beginning and end of help text
- Removed double new-line between option groups for consistency. Now all
sections have the same number of new-lines
- Switched usage and description order
- Only show "Usage"-string if no App name is present. This provides
better readability
- Made categories (Options, Positionals, ...) capital
- Changed `ConfigBase::to_config` to correctly process capital
"OPTIONS"-group (only affects descriptions of the config file, not a
breaking change)
- Added a paragraph formatter function `streamOutAsParagraph` to
StringTools.hpp
- Made "description" a paragraph block with correct, word respecting
line wrapping and indentation (using the new paragraph formatter
function)
- Made the footer a paragraph block with correct, word respecting line
wrapping and indentation
- Updated documentation for `column_width_` to make it more clear
- Added new member: `right_column_width_`, added getter and setter for
`right_column_width_`
- Added new member: `description_paragraph_width_`, added getter and
setter for `description_paragraph_width_`
- Added new member: `footer_paragraph_width_`, added getter and setter
for `footer_paragraph_width_ `
- Positionals description are now formatted as paragraph with correct,
word respecting line wrapping
- Options description are now formatted as paragraph with correct, word
respecting line wrapping
- Short and long options/flags/names are now correctly formatted to
always be at the right position (also for subcommand options/flags)
- Short and long options/flags/names column widths scale linearly with
the `column_width_` attribute to better adapt to different
`column_width_` sizes
- Merged PR #860
## What's planned for the future?
- I'm thinking of better formatting the options of flags (like REQUIRED,
TEXT, INT, ...) and make them also in a seperate column. This way they
would also always be at the same position. However I decided against it
for this PR, since I wanted them to be as close as possible to the
actual flag. With my implementation it is quite easy to add this change
in the future.
- Subcommands: I'm planning on better formatting the Subcommands. With
this PR only the short and long flags/options of subcommands are better
formatted (like it is with the main flags, see images down below).
- Maybe implement a different way to display expected data type options
(TEXT, INT, ...). For example: `--file-name=<TEXT>` for long flags only
and if `disable_flag_override_` is false.
- Maybe add something like this:
https://github.com/CLIUtils/CLI11/issues/554
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Philip Top <phlptp@gmail.com>
From #1067, there is a bit of ambiguity in the handling of config file
with vector input and multiple consecutive parameters. For example
```toml
option1=[3,4,5]
option1=[4,5,6]
```
Currently this is handled as if it were
```toml
option1=[3,4,5,4,5,6]
```
But this could be confusing in the case where the input was referring to
a vector of vectors.
This PR adds a separator in the sequence to separate the vector so they
are two vectors of 3 elements each.
Will need to verify if this change has other side effects. It is a
pretty unusual situation.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This is the next phase of the fuzzer. It runs a round trip and makes
sure that the config files generated by the app will load into the same
results, to test full round trip on the config files.
Issues fixed
- fix a bug in the string escape code caught by initial round trip tests
- resolve inconsistencies in handling of {} for empty vector indication
between config and cli parsing
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Take the configurability of an option into account when determining
ambiguous names and conflicts.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This PR adds a mechanism to hide groups but have their options visible
as part of the parent.
This works for option group names starting with a '+'
for example
```
CLI::App app;
bool flag = false;
std::optional<bool> optional_flag = std::nullopt;
app.add_option("--tester");
auto *m1=app.add_option_group("+tester");
m1->add_option("--flag", flag, "description");
m1->add_option("--optional_flag", optional_flag, "description");
CLI11_PARSE(app,argc, argv);
```
will produce help as
```txt
Options:
-h,--help Print this help message and exit
--tester
--flag BOOLEAN description
--optional_flag BOOLEAN description
```
instead of
```
Options:
-h,--help Print this help message and exit
--tester
[Option Group: tester]
Options:
--flag BOOLEAN description
--optional_flag BOOLEAN description
```
Fixes issue #1034 and a few other past issues or questions
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Added include-what-you-use pragmas to:
* let IWYU point users to the main include file CLI/CLI.hpp
* tell IWYU that CLI/CLI.hpp is the main exporting header.
This should fix#816
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Philip Top <phlptp@gmail.com>
get_subcommand when used for parsing config files, was throwing and
catching as part of control flow and expected operation, this resulting
in a performance hit in select cases. A get_subcommand_no_throw was
added to resolve this issue.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
The extra leading `"CLI/"` part of include directives prevents the
inclusion of `CLI.hpp` from a relative directory without an extra `-I`
or `/I` compiler directive, and makes it harder to make CLI11 part of a
larger codebase.
This is a regression of #475.
Fix an issue with environmental variable parsing in option_groups, and
an issue with remaining in config files.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Correct an anomaly when using config file processing with a default. In
this case the count always shows 1 even if the default file were not
actually used. This caused some issues in some applications and was a
change from previous versions.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Update and test with some newer compilers and cmake versions
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Add escaping to quoted strings, differentiate between literal and
regular strings.
The goal is to make string processing in config files as close as
possible to toml standards. This means handing escape sequences
including unicode, and differentiating between literal strings and
regular strings in files and when splitting the command line. Also
allowing variable names in the files to be quoted.
This PR gets partway there. Removes some hacks from the previous PR to
deal with unusual option names and replaces with the quoted names.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This adds a round trip test for config file generation to the fuzzer.
(the next step after this PR will be a fuzzer that verifies that the
round trip actually matches the results.
This change ended up requiring quite a few minor changes to fix the
ambiguities between the config file generation and config file reader.
1). There was a number of potential conflicts between positional names
and regular option names that could be triggered in config files, this
required a number of additional checks on the positional naming to
ensure no conflicts.
2). flag options with disable flag override can produce output results
that are not valid by themselves, resolving this required flag input to
be able to handle an array and output the original value set of results.
3). strings with non-printable characters could cause all sorts of chaos
in the config files. This was resolved by generating a binary string
conversion format and handling multiline comments and characters, and
handling escaped characters. Note; I think a better solution is to move
to fully supporting string formatting and escaping along with the binary
strings from TOML now that TOML 1.0 is finalized. That will not be this
PR though, maybe the next one.
4). Lot of ambiguities and edge cases in the string splitter, this was
reworked
5). handling of comments was not done well, especially comment characters in the
name of the option which is allowed.
6). non printable characters in the option naming. This would be weird
in practice but it also cause some big holes in the config file
generation, so the restricted character set for option naming was
expanded. (don't allow spaces or control characters).
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Add some additional tests to try to get test coverage back to 100%
refactor the positional parsing to reduce code duplication
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Fixes#845 as discussed.
Comparing the two approaches of getting `argv`:
1. The "old" way, through `CLI::argv()`:
✔️ Works automatically and almost everywhere
❌ Small abstraction overhead on macOS
❌ Does not work in weird edge-cases such as missing `/proc`
2. This PR, through `app.ensure_utf8`:
✔️ True zero-overhead abstraction: you don't pay for what you don't use
✔️ Less moving parts than the "old" approach, probably can't be broken
❌ Requires extra code to be written by the user (which is sad because
ideally everyone should use this by default)
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
use it for the default in `set_config` and simplify and add more
flexibility to the the config processing, and potentially in other
options as well.
The reverse policy returns a vector but in reversed order from normal.
This is what we want in the config processing
Inspired by #862, and updated with recent code changes.
---------
Co-authored-by: Volker Christian <me@vchrist.at>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
about an out of range string conversion not being caught properly. This
commit changes the logic from an exception to using errno and a
non-throwing alternative.
Issue detected in
https://github.com/CLIUtils/CLI11/actions/runs/5500247554/jobs/10023032108
The problem string was set up as a test.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Do not check environment variables if a subcommand has not been
triggered.
Fixes#903
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Fixes#890
Add parsing of environmental variables when supplied for the config file
option.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* update the storage of remaining argument for config files, so that they work through the `remaining_for_passthrough` with values when parsed a second time.
* style: pre-commit.ci fixes
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Add unicode support tests
* Add unicode parse tests
* Implement #14
* Slim down Windows.h
* Fix documentation comments
* Fix clang-tidy and cpplint
* Update README
* Fix clang-tidy
* Fix to_path not being available on linux
* Add roundtrip encoding tests
* style: pre-commit.ci fixes
* Fix pre-commit.ci
* Fix codacy
* Exclude parse_unicode which should not contain a newline from pre-commit
* Remove a test which breaks CI
* Fix build in CI
* Replace broken execute_with tests
* Fix wide string conversions on all systems
* Fix system args on apple
* style: pre-commit.ci fixes
* Fix some includes
* Fix wrong size calculation and comments
* Add guards around codecvt
* Fix _Pragma not recognized on MSVC
* Fix bad macro check
* Fix include
* Fix narrow and widen when codecvt is missing
* Fix some weird bug in old MSVC
* Add dependent applications to meson-build
* Fix precompilation
* Fix lint
* Fix coverage
* Update README
* style: pre-commit.ci fixes
* Fix lint
* Fix coverage
* Fix optional braces offending clang
* Remove copied comments from Windows.h
* Suppress flawfinder detects
* Fix cmake config tests failing because of a missing lib
* chore: update copyright on new files to 2023
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
* style: pre-commit.ci fixes
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Henry Schreiner <henryschreineriii@gmail.com>
* Allow using dot notation for subcommand arguments such as --sub1.field
* add tests for dot notation for subcommands
* style: pre-commit.ci fixes
* add test for short form arguments in dot notation
* style: pre-commit.ci fixes
* add _pre_parse_callback_ support using dot notation
* style: pre-commit.ci fixes
* update cmake tests to include 3.24
* change line endings
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* fix: show newline before footer only if footer is set and not empty
* feat: added usage message replacement feature
* fix: tests corrected for new help message formatting
* Using ADL everywhere for lexical_cast
* Fixes in docs
* Add a test for old extension mechanism
* style: pre-commit.ci fixes
* Make gcc happy
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Philip Top <phlptp@gmail.com>
* tweak the parsing of files for flags with disable_flag_override, basically allow true as a valid value to be interpreted as the default in that case
* style: pre-commit.ci fixes
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* the get_option_group definition was missing from the splitting of the calls.
* style: pre-commit.ci fixes
* the get_option_group definition was missing from the splitting of the calls.
* style: pre-commit.ci fixes
* add test for missing function
* style: pre-commit.ci fixes
* add test for get_option_group
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* add some more tests for coverage and fix some confusing error situations with config files.
* style: pre-commit.ci fixes
* fix warning
* ci: fix coverage
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Henry Schreiner <HenrySchreinerIII@gmail.com>
- Add C11_COMPILE cmake option that creates a static lib instead of header-only
- Add C11_INLINE macro that depends on C11_COMPILE
- Split App.hpp into App.hpp and impl/App_inl.hpp
- Add App.cpp that compiles App_inl.hpp into an object file
- CMake modifications to handle impl headers differently for sinlge-header, headers-only, and compiled versions
