Per one of the diff comments on #1597:
> I used to use a tuple for constant things, but nowadays I'd actually
> prefer a list as a tuple is something more heterogeneous (i.e. it
> makes sense to have a `(x, y)` point as a tuple, but a list of points
> would be a list).
> At some point I should probably change it to a list everywhere 😉
This approach is not as weird in Python and still works.
DownloadTarget.OpenDownload has been renamed to OpenFileDownloadTarget,
since OpenDownloadDownloadTarget didn't look as nice.
This parameter replaces the filename and fileobj parameters. This makes
it easier to add more download targets, since only one may be "chosen".
With the OPEN_DOWNLOAD special case added, handling of filename got a
bit ugly, since it may be either None, OPEN_DOWNLOAD or a str with the
file path, and we had to make sure only one target was chosen.
With the new target enum, this handling can be simplified and we
automatically get the guarantee that only one target is chosen.
The test was failing because of two reasons:
First, the old code had filename questions in DownloadManager.get and
DownloadManager.fetch which were almost identical, thus the part in
DownloadManager.get was removed in an earlier commit. All filename
asking is now done by DownloadManager.fetch. The good part is code
deduplication, the bad part is slightly modified behavior: The new code
doesn't wait for a filename to start the download, instead it tries to
fill the buffer immediately. This made the test fail because qute:// has
no registered handler, so in order for the test to pass now, the "no
crash" part is not enough, we also need to expect the "No handler"
error.
Secondly, and a rather rare (race) condition was the handling of errors
in the DownloadItem. If an error occured after the registration of
self.on_reply_error as error handler and before the check
reply.error() != QNetworkReply.NoError
at the end of the function, the error signal would be emitted twice:
Once by _die() (called by on_reply_error), and once by the init_reply
function directly (in the last if block). This lead to duplicated error
messages. This is also explained in a comment in the file (with small
"stack traces").
This way, all temporary downloads will end up in the same directory and
everything is cleaned up at program exit, not when the corresponding
window is closed.
The only thing which differs from url() is that it got set immediately
after openurl() was called, which might or might not have improved
something.
Let's see if things still work the same without it.
As mentioned here:
e4b0b7fffd (r70002693)
It makes no sense to add a backend-specific run_webaction method to
AbstractTab - better to just access _widget directly in this one place
instead of adding something backend-speficic to the API.
By default, the AbstractTab object got the focus, which means things
like key events passed to it didn't actually get passed through to the
web view, causing these tests to fail:
tests/end2end/features/test_keyinput_bdd.py::test_forwarding_all_keys
tests/end2end/features/test_keyinput_bdd.py::test_forwarding_special_keys
Now we make sure the real underlying WebView always gets the keyboard
focus.
While we need to set it from the outside (from AbstractTab) this still
is not considered public API for the rest of the code, so let's make it
private.
It was implemented as a member of the Completer object, but registered
to the CompletionWidget. This led to the weird scenario where self was
actually a CompletionWidget, even though it was declared in Completer.
- Use the whole home row
- Produce easier chords
- 2-key chords alternate hands first (f[aoeui][dhtns])
- This is indeed a poor patch for easier chords as it should work on
every home-row, but there are not that much homerows around.
When a redirect occurs, the item is saved in history with a -r suffix
now. When opening qutebrowser that's picked up and the item is hidden
from completion.
- Rename HistoryEntry to Entry
- Move history line parsing from WebHistory.async_read to Entry.from_str
- Improve errors for invalid history lines
- Pass history directory/filename from the outside to WebHistory
- Clear temp history after reading it when async_read is done
This makes it possible to switch to an alternative implementation if
there are weird issues like #1568. Some users might also prefer the
slightly better performance over more accurate hints.
This won't hurt and will help with some poorly formatted sites
including blank spaces around the url (e.g. the Previous link in a
dashboard make with CDash 2.0.2).
As described in [1], the naming of some variables has become
inconsistent with the original code and even docstrings. This commit
corrects some of these problems, with the following terminology:
- hint text: informative message (see HINT_TEXTS)
- hint string: the text displayed on the hint (as instance of str)
- hint label: the element representing the hint, added to the DOM
- hint: too abstract, sensibly used only in docstrings to refer to the
"visual result"
This commit amends b89e0f8803 and
8873aba09f.
[1] https://github.com/The-Compiler/qutebrowser/pull/1178#issuecomment-178795190
Both set mode=None, then later checked if mode == 'None' and set it to
'normal'.
This reduces the function complexity just enough that pylint will stop
complaining.
Load all special keystrings (e.g. <ctrl-a>) into memory as lowercase,
and automatically lowercase any special keystring given to bind/unbind.
This prevents <ctrl-a> and <Ctrl-A> from being treated differently.
Resolves#816.
Also resolves#1544 (dupe).
On OS X, sometimes the userscript exited before the QSocketNotifier was
triggered. By doing a final read when cleaning up we make sure we don't
miss anything.
Fixes#1533.
Replace the setting ui.show-keyhints with ui.keyhint-blacklist, which
is a list of globs for keychains that shouldn't be hinted. This allows
users to prevent showing keyhints for keychains they already know.
keyhint-blacklist='*' is equivalent to show-keyhints=False.
Resolves#1515.
If a user knows the keychain and can type it quickly, we shouldn't
annoy them with a popup. Only show the keyhint if the user doesn't
complete their keychain in 500ms.
The isVisible() check in the tests is somewhat invalid now because it
is never immediately visible and I don't want to add a delay to unit
tests. I added a check that text() is not set for one test that was
only checking isVisible().
Addresses part of #1515.
This supports things like :hint all spawn -v echo as '-v echo' will be
passed as a single unit to spawn rather than -v being interpreted as a
flag for :hint.
Resolves#797.
Note that, while `:hint --rapid all spawn -v` echo works,
`:hint all --rapid spawn -v echo` does not (this did not work before
either).
Instead of creating a new guiprocess manually, just pass the args along
to the spawn command so it can accept args like -v.
Addresses part of #797 by allowing `hint -- all spawn -v echo`.
`hint all spawn -v echo` is still not supported.
When we are in rapid mode with only one link, after following the hint, fire()
called filter_hints(None) to display all hints again. Then filter_hints tried
to follow that link, fire() tried to show all again, etc., leading to a
RecursionError.
Fixes#1513.
A test will be added via #1510.
The check `key.startswith('<') and key.endswith('>') is repeated many
times in code to check for a special key. Replace all these with a call
to the same function.
Currently, the keyhint window is shown even if the keystring matches no
possible bindings. This causes an empty keyhint window to hang around
after entering hinting mode.
Instead, the window is now hidden if no bindings match the current
keystring.
Resolves#1507.
When showing the currently bound key in the misc column for command
completion, if the command has multiple bindings, show special bindings
(e.g. <ctrl-a>) after non-special bindings.
- Add a space after the comman for multiple binding suggestions.
- Use defaultdict(list) instead of defaultdict(lambda: [])
- Move the pylint comment back to the top of the class
Since we're not using those functions as argparse callbacks anymore, we
can write a normal function instead of factories, which simplifies
things a lot.
This means:
- An annotation like (int, str) is now typing.Union[int, str].
- utils.typing got expanded so it acts like the real typing.py, with
issubclass() working properly with typing.Union and __union_params__
being set.
- A literal string doesn't exist anymore as annotation, instead
@cmdutils.argument now has a 'choices' argument which can be used like
@cmdutils.argument('arg', choices=['val1', 'val2']).
- Argument validating/converting is now entirely handled by
argparser.type_conv instead of relying on python's argparse, i.e.
type/choices is now not passed to argparse anymore.
TestArgument didn't clear the globals as the fixture was inside
TestRegister.
This means test_run_vulture failed in funny ways because run_vulture.py
generated a whitelist containing "<locals>" for commands:
tests/unit/scripts/test_run_vulture.py:55: in run
return run_vulture.run([str(e.basename) for e in files])
scripts/dev/run_vulture.py:146: in run
vult.scavenge(files + [whitelist_file.name])
.tox/py35/lib/python3.5/site-packages/vulture.py:107: in scavenge
self.scan(module_string)
.tox/py35/lib/python3.5/site-packages/vulture.py:75: in scan
node = ast.parse(node_string, filename=self.file)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
source = 'qutebrowser.browser.commands.CommandDispatcher.buffer\nqutebrowser.misc.savemanager.SaveManager.save_command\nqutebro...iidoc.UsageFormatter._get_default_metavar_for_positional\nscripts.dev.src2asciidoc.UsageFormatter._metavar_formatter\n'
filename = '/tmp/tmp_ein2umn', mode = 'exec'
def parse(source, filename='<unknown>', mode='exec'):
"""
Parse the source into an AST node.
Equivalent to compile(source, filename, mode, PyCF_ONLY_AST).
"""
> return compile(source, filename, mode, PyCF_ONLY_AST)
E File "/tmp/tmp_ein2umn", line 16
E test_cmdutils.TestArgument.test_wrong_order.<locals>.fun
E ^
E SyntaxError: invalid syntax
Previously, negative bottom padding was used (probably to compensate
for a trailing <br>). With the table format, this is no longer
necessary and causes the last line to be drawn too low.
This allows webkit to color links that are clicked on but never rendered as
visited too. It also means if you get redirected from eg http://site.com to
http://site.com/ you have essentially duplicates in your history. This makes
the history completion a bit noisier. I suppose normalising paths before
checking for duplicates might help. Also note that otter has an isTypedIn flag
which might be used for dealing with this.
Now adds a url to browser history once we have connected and got enough data
to start rendering the page. The previous approach saved urls as soon as
navigation was initiated, so upon encountering a redirect the final url wasn't
saved.
Using layout started rather than load finished means that pages whose contents
manage to load minus one troublesome asset will still be saved.
Adds a title to the HistoryEntry class and includes it in the serialization
stuff. Not currently set from anywhere.
Not sure if anything more needs to be done to support non-ascii characters.
Everything works fine for me with unicode chars in url and title but
everything in my stack is utf-8.
Don't show special keys in the keyhint window as these currently cannot
be part of keychains.
Use a rounded border on the top-right corner and square on the rest.
This failed because dict.get('level') returned None with no level
parameter, and the subsequent [0] raised:
Traceback (most recent call last):
File ".../qutebrowser/utils/utils.py", line 624, in wrapper
return func(*args, **kwargs)
File ".../qutebrowser/browser/network/networkmanager.py", line 445, in createRequest
op, req, outgoing_data)
File ".../qutebrowser/browser/network/qutescheme.py", line 107, in createRequest
data = handler(self._win_id, request)
File ".../qutebrowser/browser/network/qutescheme.py", line 189, in qute_log
level = urllib.parse.parse_qs(request.url().query()).get('level')[0]
TypeError: 'NoneType' object is not subscriptable
With the changes to get rid of colorlog, we broke this, as e.g. {green}
was an undefined key for the vanilla logging.Formatter used for the
in-RAM log. Now we instead use a ColoredFormatter with colors turned
off.
From code review:
- escape all strings used in the keyhint html
- read the prefix color each time the hint is shown
- use show/hide instead of setVisible
- clean up pylint/flake8 errors
- use CssColor instead of QssColor for keyhint.fg.suffix
- add some padding to the keyhint popup
When the current keystring is a partial match for one or more bindings,
show the possible bindings in a small overlay.
The overlay is partially transparent by default, but the background
color is configurable as ui->keystring.bg.
- Add log.LOG_LEVELS to map names to levels (instead of using
logging._levelToName)
- Test that log pages do not contain messages below the requested level
- Use pythons urllib.parse.parse_qs instead of Qt's UrlQuery
- Document tab, bg, window args for :messages
- Clean up style
This adds a 'level' query parameter to qute://log and qute://plainlog.
For example, qute://log?level=warning will show an html page containing
log entries with severity warning or greater.
If the query is omitted, the original behavior of qute://log is
preserved.
:messages [level] is a command that opens qute://log?level=<level>.
By default, level defaults to 'error' as an easy way to see missed
error messages.
colorlog was problematic for various reasons:
- Not commonly packaged for Linux distributions
- Calling colorama.init() automatically on import
- Not supporting {foo} log formatting
- Not supporting an easy way to turn colors off
Instead we now do the log coloring by hand, which is simpler and means
everyone will have colored logs.
The output failed the tests as it was parsed as invalid. Indenting helps
with that.
This also simplifies things a bit by having a _log_stack function and
shows the type of stack we're printing.
Before we used a {'nargs': '*'} annotation for the respective argument
to tell qutebrowser it's optional for the commandline. Now we instead
use a star_args_optional argument for @cmdutils.register as a first step
towards freeing up argument annotations for PEP 484.
See #637.
The CompletionItemDelegate gets reused by Qt for various items in the
completion. Every time _get_textdoc() was called we created a new
QTextDocument, but since it has a long-living parent set (the delegate)
the old one was never actually garbage collected.
We now explicitly delete the old QTextDocument as it's not needed
anymore by either Qt or Python.
See #1476.
len(self.downloads) is already the index of the item in the download
list, this should be used for beginInsertRows(). The +1 is only for the
human readable part.
Issue #846
.rowCount() returns all downloads, even the finished ones that have not
yet been removed from the list. For confirming the quit event, we should
only consider downloads that are still running.
In on_config_changed, we now ignore FileNotFoundError as that's a common
occurence and not something worth logging.
In case of other OSError's we now also log the exact error message.
Issue 1413
This happens when the networkmanager is used by something that has no
tab_id, like the generic DownloadManager. In this case, we should just
skip the webview connection (as it makes no sense) instead of crashing
(which is the last thing we want to do).
All commands will be offered as completions for the <command> argument
of :bind.
Due to the way completers parse the command line, the following
bind --mode caret j
will throw off completions as 'caret' is treated as a positional arg in
terms of the argument count for completions.
In the above example, completion will be triggered for 'j' and no
completions will be given for the actual command.
bind --mode=caret j will complete correctly, though completions are not
filtered by the given mode.
I attempted an approach to filter the commands based on the mode but it
ended up being messy and flaky.
The <command> arg is now optional. If omitted, :bind prints the current
binding as a message. If --mode is given, the binding for that mode is
printed.
Allow a variable amount of whitespace for rgb, rgba, hsv, and hsva
strings in the config.
Previously only 'rgb(0, 0, 0)' was allowed. Now things like
'rgb(0,0,0)' are permitted.
The repeated 3-digit segments of the regexes were separated out to
reduce repetition and line length.
If the process emitted error() and then finished(), we already set
self._filepath to None and did other cleanup.
Instead we do the file reading inside _cleanup and call that from
on_process_error and on_process_finished.
