I want to use the SFTP protocol to create folders on the remote SFTP server, without specifying the curl option. The option created on the FTP protocol is "CURLOPT_FTP_CREATE_MISSING_DIRS".
Do you have any corresponding SFTP options?
Thanks for any help.

For example device()method can also return false not just string.

Here is the method declaration:

/**
     * Get the device name.
     *
     * @param  string $userAgent
     * @return string
     */
    public function device($userAgent = null)
    {
        // Get device rules
        $rules = $this->mergeRules(
            static::$additionalDevices, // NEW
            s

• Mojolicious version: 8.29
• Perl version: 5.26.1
• Operating system: High Sierra

When testing a JSON value like { "fruit": "lemon" } with ...->json_is('/animal', 'bat') we correctly get a failure, but the diagnostic message complains that undef doesn't equal 'bat' which isn't quite right. The value isn't even undef, it doesn't exist. Same applies to json_like.

Many Browsers (and Brands) detected by DeviceDetector have vague names that make it hard to find out more about them (or find a logo for them). Others have names, that are definitly not unique (START Internet Browser). Thankfully, many of them have a comment in the YAML file pointing to a Website with more information (or a app store entry).
e.g.
https://github.com/matomo-org/device-detector/

• Filestack
• Google-Ads-Overview Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2272.118 Safari/537.36
• Google-Ads-Overview Mozilla/5.0 (Linux; U; Android 6.0.1; generic) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2272.118 Version/4.0 Mobile Safari/537.36
• Google-Ads-Overview Mozilla/5.0 (Linux; U; Android 2.3.4; generic) AppleW

I think its better to add osName enumsto docs.
By now just two of os name enums are specified in docs:
returns os name (e.g Windows, Android)
In my case I should send to server specific enums based on OS and need to know exactly what os names should I expected from this method to add to my local switch.

Example test case:

...
{
	title:    "Pingdom",
	ua:       "Pingdom.com_bot_version_1.4_(http://www.pingdom.com)/",
	expected: "Pingdom:1.4 Browser:Pingdom-1.4 Bot:true Mobile:false",
},

Test     Pingdom
got:     "Browser:Pingdom.com_bot_version_1.4_(http: Bot:true Mobile:false"
expected "Pingdom:1.4 Browser:Pingdom-1.4 Bot:true Mobile:false"

Not really importa

Might be good to add this to the documentation :

• user-agent strings provided is not "random" because user-agents.json.gz contains browser fingerprints not user-agents strings so it will give you a representation of what is most used at the time period depending of the version of the lib.
• the db is updated in full not incremental, so old UA are ventilated

For example :

Top user agen

Small observation:
When choosing an user agent, it requires selecting the user agent, followed by clicking "Apply".
When resetting the user agent (using the system's user agent), pushing "Reset" will apply it automatically.

It would be more consistent to always require "Apply" in order to apply any user agent changes.

The types returned from each method/attribute are really confusing - for example here:

> platform = ua.device.platform
=> "iPhone"

where ua.device.platform actually returns a AgentOrange::Platform object. Am I missing something? Documenting with YARD would be awesome, too...

http://rubydoc.info/gems/ag

#47 introduced Proxy filtering as an optional parameter in the RequestProxy constructor.
It would be great to extend #5 to support that!