Once before installing/updating to prepare _zmodule_xargs, and once
after modules are updated, so functions and scripts can be found inside
them. Installation of Zim from scratch was failing because all modules
are empty at first.
to autoloads the functions and sources the scripts, instead of executing
zimfw during startup, and having it always figuring out what do to on
the fly.
This takes out the worry about zimfw interfering with the startup time,
and allows room to add more features to it. So, zstyle was replaced by a
custom zmodule function to define the modules, with the extra ability of
allowing users to set custom fpath paths, autoloaded functions and
sourced scripts per module.
because it was being processed as the beginning of a escape sequence.
Using `print -R` fixes that. Probably a good idea to use it when
printing other messages that contain externally-generated output.
Also moved the templates out of this repository, and into the
zimfw/install repo.
This is a second big change after introducing the plugin mechanism. This
makes installation and upgrading of Zim straightforward. Maybe the most
important aspect of having the script in a single file is not having to
manage "git repos inside git repos" (see #297), since the single file
exists by itself and is not version-controlled (with git).
I've implemented a two-stage sourcing of the file, so most of the file
is only sourced when needed (namely when calling `zimfw` with any action
other than `login-init`). The two-stage process is designed to avoid
compromising the startup speed, which is our top priority.
In an effort to help making the script maintainable, I've broken it into
small ERB templates. This also adds the ability to pre-process the Zsh
code with Ruby code. To build the script, use `make`.
Only 5% (18/342) of the themes listed under [unixorn/awesome-zsh-plugins]
are actually compatible with prompinit. Of these, [clean] also allows
being sourced directly. On the other hand, 3 others are prezto themes.
promptinit would be useful for who wants to try many themes without the
need to restart their shell session. And must be many many, so
"brute-force" starting a new shell to experiment each new theme would be
a burden! Even the cleanup feature of promptinit is still incomplete, so
you eventually get a messy prompt after trying many with it. And that's
not even a everyday use case of the average Zsh user.
So prompinit it not widely supported out there, and also not very useful
for the everyday let-me-use-my-beloved-and-carefully-customized-prompt-during-the-whole-shell-session-pleasee
scenario. It's also faster and simpler to directly just source the prompt
theme to be used, not even having to autoload promptinit and let it scan
all the others themes in fpath that won't be used.
And the Zim "philosophy" is to use fast and simple solutions.
So here we go.
Fixes#325
[unixorn/awesome-zsh-plugins]: e226f3de04/README.md (themes)
[clean]: https://github.com/BrandonRoehl/zsh-clean
so going into their ${ZIM_HOME}/modules/foo directory would be like
still being inside the Zim repo, in ${ZIM_HOME}. Don't try to update
them in this case.
This was supposed to be working before, but my ${ZIM_HOME} was not a
repo when I was still developing this locally.
Also change formatting of the settings. Using `<em></em>` to prevent
URL autolinking in some cases, and to add emphasis where `*foo*` would
be ambiguous for the Markdown format.
Copy was also slightly improved, hopefully for better clarity.
This is a major change, where Zsh modules/plugins are not git submodules
in the Zim repo anymore, but customized and installed separately as
individual repositories. The discussion about this started more than 2
years ago in #88. Closes#299.
This will allow contributors' modules to live in their own repositories.
Closes#33, closes#138, closes#262, closes#277, closes#281.
Some discussion topics that I think are worth considering before merging
this:
- [ ] Reduce the Zim "core" to a single file?
- [ ] Simplify installation? With an installation script? (See #182)
- [ ] Put the configuration into `.zshrc` instead of a separate `.zimrc`?
(See #288)
- [ ] Rerun the Eriner/zsh-framework-benchmark?
I suggest we create individual GitHub issues/PRs to start the separate
discussions.
The current code has what, up to this point, I considered to be the best
balance between simplicity, execution speed and number of files.
One measured decision was to make the initialization of modules depend
only on the `':zim' modules` style, keeping it as fast as possible.
The `':zim:module' module` style is used to install, update and clean
the modules, all operations that happen after the user got his
as-blazing-fast-possible shell prompt.
Even though I didn't care much about making install or update fast,
`xargs` has a nice feature of allowing commands to be executed in
parallel with `-P`. I took advantage of that.
I've also worked on making the `zimfw` utility give the user some nice
(while still minimalistic) output. Also I'm suggesting this as the new
name for the `zmanage` tool, since `zimfw` does not shadow the `zim`
wiki tool.
I strongly recommend you install this from scratch in a separate
directory, instead of checking out `develop` in your current Zim
installation repo.
Stick with the following style:
* Header 1 with the module name is in `lowercase`.
* Other headers are in `Sentence case`. Common header names that should
be consistently used are `Aliases`, `Contributing`, `Functions`,
`Settings`, and `Zsh options`.
* The names `Zim` and `Zsh` always appear capitalized, even in the
middle of sentences.
* Prefer
print 'code indented with 4 spaces'
instead of
```zsh
print 'code fenced by lines with three back-ticks'
print 'unless you want syntax highlighting'
```
Prefer `setopt NO_FOO` instead of `unsetopt FOO`, as former is easier to
document in the README.
Stick with the following style for the README:
* Header 1 with the module name is in `lowercase`.
* Other headers are in `Sentence case`. Common header names that should
be consistently used are `Aliases`, `Contributing`, `Functions`,
`Settings`, and `Zsh options`.
as they are not required (not even recommended) to be set along with
`SHARE_HISTORY`. See zshoptions(1) on `SHARE_HISTORY`:
> This option ... also causes your typed commands to be appended to the
> history file (the latter is like specifying `INC_APPEND_HISTORY`,
> which should be turned off if this option is in effect). The history
> lines are also output with timestamps ala `EXTENDED_HISTORY` ...
Also update copy in comments and in the README. Stick with the following
style for the README:
* Header 1 with the module name is in `lowercase`.
* Other headers are in `Sentence case`. Common header names that should
be consistently used are `Aliases`, `Functions`, `Settings`, and
`Zsh options`.
* The names `Zim` and `Zsh` always appear capitalized, even in the
middle of sentences.
Closes#313
Add a first step to explicitly start the Zsh shell, as some issues were
reported before because users were skipping this. See #214 for example.
Also mention `~/.zlogin` as part of the uninstalling process.
Use the following formatting styles:
* Headers >1 are in `Sentence case`.
* The names `Zim` and `Zsh` always appear capitalized, even in the
middle of sentences.
* Prefer
code indented with 4 spaces
instead of
```
code fenced by lines with three back-ticks
```
Closes#315
namely, the `gfu` alias, the `git-branch-delete-interactive` function,
and a couple of commands that are shadowed.
Command `gm` was reported in #59, and `grc` was reported in #139.
Prevent main repo `git status` reporting submodules having new files.
New files are usually zcompiled zsh files.
There's no way to .gitignore from main repo specific files in submodules
Closes#297
Current code will create an archive in the current directory with:
% archive ../test.tar.gz test.*
or will complain that the following archive doesn't exist in the current
directory (given it actually exists in the parent one):
% unarchive ../test.tar.gz
Fix that and allow archives in any directory.
Other changes:
* Use `<required_param>` instead of `[required_param]` in the usage text
* Don't explictly check if archive exists in `unarchive`, but let the
respective tool fail with its own message
Closes#312
My fault, I removed them in commit 78611457, but they are necessary.
Without them, we are getting the following error when one of the
terminfo[] values is empty:
init.zsh:14: bad set of key/value pairs for associative array
Fixes#310
Update installation instructions in README.md to use cat. Also add blank
lines at the end (instead of beginning) of template files, since they're
prepended (not appended) to existing files.
See difference of output between print and cat (zlogin having a blank
like at the end):
% print -rn "$(<zlogin)$(<test)"
#
# User configuration sourced by login shells
#
# Initialize zim
[[ -s ${ZIM_HOME}/login_init.zsh ]] && source ${ZIM_HOME}/login_init.zsh# Hello world
% cat zlogin test
#
# User configuration sourced by login shells
#
# Initialize zim
[[ -s ${ZIM_HOME}/login_init.zsh ]] && source ${ZIM_HOME}/login_init.zsh
# Hello world
Fixes#94. Fixes#280. Closes#300
In double quotes, array elements are put into separate words when
using `"${(@)array}"` or `"${array[@]}"`. See zshexpn(1).
Also according to the Zsh documentation, these forms preserve empty
elements of the array.
instead of aliasing `ls` to `ls -G`.
CLICOLOR is detected not only by `ls`, but by `tree` too (starting with
version 1.8.0). See http://mama.indstate.edu/users/ice/tree/changes.html
It's probably more widely used by other tools too.
not just one directory.
Also use shorter versions of the `tar` parameters, since we were using a
mixture of the short and long ones among `archive` and `unarchive`.
About not using `$` inside `(( ))`, this is what the the section
ARITHMETIC EVALUATION in zshmisc(1) says:
> Named parameters and subscripted arrays can be referenced by name
> within an arithmetic expression without using the parameter expansion
> syntax.
And according to http://www.bash2zsh.com/zsh_refcard/refcard.pdf:
> `var` (does not require `$` in front unless some substitution e.g.
> `${#var}` is needed, `$` is error if `var` is to be modified)
Closes#308
Add missing local variable set.
Replace `>/dev/null 2>&1` by `&>/dev/null`.
Check return value of zstyle (if zero), instead of resulting string
value (if non-empty).
I think this change results in the _expected_ behavior. Users can always
override this by overriding the zstyle, as that is a benefit of using
zstyles.
Ref: https://www.zsh.org/mla/users/2015/msg00467.html
and `which _ssh_hosts`
Fixes#46. Closes#294.
based on https://stackoverflow.com/a/48509425/2654518, which is based on
http://rabexc.org/posts/pitfalls-of-ssh-agents
Current code has a few issues: depends on `SSH_AUTH_SOCK` and
`SSH_AGENT_PID` env variables, which might not be available in every
shell session; and tries to create a new socket for agent-forwarding by
checking `SSH_AUTH_SOCKET` instead of `SSH_AUTH_SOCK`.
Also, it's safer to create the env file with 066 mode and in the user
home directory.
About not using `$` inside `(( ))`, this is what the the section
ARITHMETIC EVALUATION in zshmisc(1) says:
> Named parameters and subscripted arrays can be referenced by name
> within an arithmetic expression without using the parameter expansion
> syntax.
And according to http://www.bash2zsh.com/zsh_refcard/refcard.pdf:
> `var` (does not require `$` in front unless some substitution e.g.
> `${#var}` is needed, `$` is error if `var` is to be modified)
Closes#292
Just binding ^[[A/^[[B does not work for some terminals, while not
binding them breaks zle widgets on other terminals. That's how crazy it
can get :- (
Fixes#293. Fixes#295.
See https://github.com/zsh-users/zsh-history-substring-search/issues/70
and comment by @molovo:
I was already binding keys _after_ `zle-line-init`, but I'm pulling
keycodes out of the `terminfo` module, which reports up
(`$terminfo[kcuu1]`) as `^[0A` instead of `^[[A`. Binding `^[[A`
manually mean up/down works with history-substring-search both
before and after `zle-line-init`, but if I bind `$terminfo[kcuu1]`
(`^[0a`), up/down only works *before* `zle-line-init`.
Closes#286
and they were mistakenly inside the Colours scope. Move them out of
there.
Still using the same `(( ${+commands[dircolors]} ))` test because it
works well with MacOS with coreutils installed without the `g` prefix,
for example.
This reverts commit 2f0243e533.
Completion module does not depend on utility module anymore to set the
zstyle list-colors. Making modules independent from each other as much
as possible is a good thing.
