non-ASCII usernames character will cause the module's plot to not work

[shun2wang]

• OS: windows 10 21H2
• locale: Chinese
• version: 2.3.6

Hi @jonathon-love

I got a lot of help request after we released 2.3.6 (but not when testing the individual patch),This issue is related to the inference of #1009 In the non-ASCII username path, jamovi cannot correctly parse the path, so the drawing in the module will not work properly, and the process of saving the analysis file .omv will prompt 'Save failed , path not found'.

Maybe with the update of R4.2 (is available now), we will be able to completely get rid of these annoying encoding characters on Windows, but now it is really a problem.

In the following example, the user name is three characters of Chinese characters '杨**' instead of the five garbled characters in the picture

thank you
shun

[jonathon-love]

thanks sw

this one i am quite puzzled by ... the code involved is entirely python, and i've never had a path encoding problem in python before ... 🤔

[jonathon-love]

could you see if it's possible to save a file without any images in it? ... if it isn't, i'd be interested to see the error message there too.

with thanks

[shun2wang]

could you see if it's possible to save a file without any images in it? ... if it isn't, i'd be interested to see the error message there too.

no, they cannot sav any .omv files, it says:'system can not find the file specified'.
I provided them with a temporary solution, which is to add an account with English letters locally and use jamovi under this account, and this problem disappears.

I have a basic understanding of python and a Chinese system environment. Maybe you can direct me to some reproducible experiments in python on windows and let me see what I can help.

thank you jonathon

[jonathon-love]

that would be awesome.

let me walk you through my reasoning on this, as you may pick up on something that i've missed.

the errors in the screenshot above are when requesting the images on line 108 here:

            resource_path = instance.get_path_to_resource(resource_id)
            with open(resource_path, 'rb') as file:
                self.set_header('Cache-Control', 'private, no-cache, must-revalidate, max-age=0')
                content = file.read()
                self.write(content)

https://github.com/jamovi/jamovi/blob/current-dev/server/jamovi/server/server.py#L107-L111

so resource_path is bad, we're thinking something has gone wrong when the paths have been created. so working backwards, we want to look at instance.get_path_to_resource(resource_id) which is:

def get_path_to_resource(self, resourceId):
    return posixpath.join(self._instance_path, resourceId)

https://github.com/jamovi/jamovi/blob/current-dev/server/jamovi/server/instance.py#L254-L255

this combines the instance path:

C:\Users\${ username }\AppData\Local\Temp\${a_generated_temp_name}\${session_id}\${instance_id}

with the resource path

${analysis_id_and_name}/resources/${ imageName }

(and we see that in the screenshot above). so the problem with the username seems to be coming from the instance_path.

working backwards, the instance path is set when the Instance object is constructed:

class Instance:

    def __init__(self, session, instance_path, instance_id, settings):

        self._session = session
        self._instance_path = instance_path

https://github.com/jamovi/jamovi/blob/current-dev/server/jamovi/server/instance.py#L69-L74

which is constructed from the session path:

        instance_path = os.path.join(self._session_path, instance_id)
        settings = await self.get_settings()
        instance = Instance(self, instance_path, instance_id, settings)

https://github.com/jamovi/jamovi/blob/current-dev/server/jamovi/server/session.py#L156-L158

which is constructed from the data path:

class Session(dict):

    def __init__(self, data_path, id):
        self._path = data_path
        self._id = id
        self._session_path = os.path.join(data_path, id)

https://github.com/jamovi/jamovi/blob/current-dev/server/jamovi/server/session.py#L52-L57

which here is called the spool path:

self._session = Session(self._spool_path, self._session_id)

https://github.com/jamovi/jamovi/blob/current-dev/server/jamovi/server/server.py#L620

which is populated using the python TemporaryFile:

        self._spool_path = conf.get('spool_path')
        if self._spool_path is None:
            self._spool = TemporaryDirectory()
            self._spool_path = self._spool.

https://github.com/jamovi/jamovi/blob/current-dev/server/jamovi/server/server.py#L507-L510

so this is really weird, because this is all sensible python code, and there aren't many pitfalls in python code.

one riddle here is that jamovi relies on creating and attaching to a shared memory segment for the data set, and this lives in the instance path, so it's weird that things based on the instance path work for some things, but not others.

one thing to be wary of, is that the corrupt user name in the terminal ... it may be correct, and may be being corrupted when the terminal displays it (and the error may be somewhere else). i've fallen for that in the past.

with thanks

[jonathon-love]

but simply running the python code:

from tempfile import TemporaryDirectory

dir = TemporaryDirectory()
print(dir.name)

on a windows machine with a problematic username, should tell us a fair bit.

[shun2wang]

Sorry for the late, I'm on vacation, but I now have time to look at this with you now.

one thing to be wary of, is that the corrupt user name in the terminal ... it may be correct, and may be being corrupted when the terminal displays it (and the error may be somewhere else). i've fallen for that in the past.

yes i have tried the TemporaryDirectory, python may be correct.

I suggest that I set up an environment like this, and then you remotely control the computer (perhaps using anydesk, and run some tests on it yourself.If you agree, I will give you a code to connect to this computer later by email

[jonathon-love]

that would be great ... i've used teamviewer in the past ... but i expect anydesk will work too. ... you have access to a computer which has this problem?

[shun2wang]

I'ii set a same environment to repreduce it for you on my computer.

[jonathon-love]

perfect. thanks. ... i gather it's only certain user names which are a problem.

[shun2wang]

i have send a code to you. waiting for you.

[sivanMargalit]

I have the same issue with non ascii character path.
Did you find a solution to this problem ?

[jonathon-love]

it's a bug in electron

electron/electron#33997

[jonathon-love]

although you won't have a problem if you use the cloud version