The value for fancyindex_css_href was not being propagated properly, so
overriding its value did not work as expected. For example the following
configuration:
http {
# ...
fancyindex_css_href "/css/global.css";
location /subdir/ {
fancyindex_css_href "/css/local.css";
}
}
would always use "global.css", even when accessing file listings under
"/subdir".
This patch fixes the issue.
This patch adds a new configuration directive fancyindex_time_format, which
accepts a strftime()-style format string used to format the timestamps in the
generated listings. The accepted format specifiers are an useful subset of
those supported by strftime(), but the libc function is not used to make the
output of the module stable and locale-independent.
This fixes issue #23.
This adds a new "fancyindex_hide_symlinks" configuration flag, which will
filter out symbolic links from the generated file listings when enabled.
Feature suggested by Peter Wemm (https://github.com/DarkHelmet433)
Fixes issue #27
If fancy indexing is used at the root of a webserver, it will still display
a "Parent Directory" link even when listing the webserver's root. This should
not happen as it outputs a useless link.
The fancyindex_footer and fancyindex_header settings need the standard Nginx
ngx_http_addition_module built into Nginx. Add a note to this effect in the
README.
Also, a warning is issued at configuration time if the addition module is not
enabled in the configuration.
Closes issue #26.
Adds a link to the nice theme designed by @TheInsomiac, which is a
great example of advanced theming using "fancyindex_header" and
"fancyindex_footer".
Closes issue #11
There is almost universal support for the CSS3
:nth-child() pseudo-selector. We're shifting the
responsibility for labeling even and odd rows to
the browser.
Set html5 meta viewport width to device-width. A mobile browser now scales up the directory listing and dir-/filenames are easier to read and click/touch.
Set html5 meta viewport width to device-width. A mobile browser now scales up the directory listing and dir-/filenames are easier to read and click/touch.
If a sorting criteria different than the default is used, carry the URL
arguments to the links pointing to other directories, so the sorting is
"remembered" when navigating across different directories.
Make table headers clickable, for changing sorting direction of the entries.
Clicking on the name of the headers selects ascending direction, clicking on
the arrow next to it, chooses descending direction. The arrow symbol, for
simplicity, is an Unicode down-arrow symbol, which all reasonably modern
browsers should display fine. Also, most text-only browsers will show the
arrow correctly when running on Unicode-compatible terminals.
When choosing different sorting criteria for the elements (with the "?C=x"
URL argument), allow also specifying a second "O=x" argument for the
direction: "?C=x&O=A" selects ascending direction, while "?C=x&O=D" chooses
descending direction.
This introduces Apache-style URL arguments to specify how to sort the
entries in the generated listings:
- Appending "?C=M" sorts by modification time (mtime).
- Appending "?C=S" sorts by file size.
- Any other (or no arguments) use the default sorting, by name.
When issuing subrequests (currently for non-builtin headers and footers),
initialize the headers_out.status field to NGX_OK.
Also, change the check of the result from ngx_http_send_header() to only
check for NGX_OK, as it is redundant to check it and also NGX_ERROR.
More exactly:
* In the requirements section, state that the module will work with
any version of Nginx newer than 0.7.x
* Updated URL to the Nginx wiki, using now wiki.nginx.org
* Specifying a custom CSS works, the README was contradicting the code.
FWIW, removed status markers from features in the file.
* Made the build instructions a bit more generic.
Unfortunately ngx_escape_uri() does not properly escape all characters
that may be problematic in URIs, so apart from doing a first pass with
ngx_escape_uri(), it is needed to do a second pass for escaping the rest
of characters.
Thanks to Steve Willing <eiji-gravion@hotmail.com> for reporting the issue.
Peter Wemm