Problem with HTML file

[matanzuckerman]

Hi @semion1956

I found a problem with the HTML. the file that the code creates is as follow (it is the same as the example you uploaded to the git)

when I'm choosing one of the config files I don't see any results. Can you please check it?

Thanks!

[tomerm]

@matanzuckerman I would think that you should provide your environment (at the very least files associated with HTML application). How otherwise we could debug the case ? Or alternatively you complain on some general phenomenon (when selecting just one experiment - does not matter which one - nothing is displayed) not tightly coupled with specific data (in other words you believe it will be reproducible on our side with our data as well) ?

Also, it is not very clear what is you expectation. Assuming you select all files (in this case just 2) , do you see a different picture ? Please observe that this UI is meant to be used for comparison between models / experiments. When you select just 1 model / file / experiment there is nothing to compare it against.

[semion1956]

@matanzuckerman 1) HTML is created and immediately shown by info.py. It contains references to js and css files, placed in TrainAndTest/info. Paths to these files are relative. If you try to launch html directly by browser you will get what you see. 2) ID-s from the right aren't "different configs", but different requests. Each request has its own unique ID, corresponds to the date and time, when it was launched. Option "infofrom" from config file can contain word "today" or string like "X days", where X is count of days, ending by today.

[semion1956]

@tomerm Check results of one request only make sense too - for example, if you want to get detail info about documents from testing set.

[tomerm]

@semion1956 I think what you are saying is that standard way to see HTML report is to run info.py. It is forbidden (not supported scenario) to open up HTML report from file system directly via web browser.

[matanzuckerman]

@tomerm The screenshot is from the git itself (and you have all the info about the env and etc for those experiments).

@semion1956, As I see it, the webbrowser need to open the html in more coherent way. The problem is that it do not do it. I think it's because I'm running the script on the server and my GUI is on the client (I'm on local and the code on the server). In this case it might try to open the html in the server and not on my local machine?

Thanks

[semion1956]

@matanzuckerman Screenshot is not from git, it is you attempt to launch html from git... Anyway, do you see at least wrong html page (as it is shown by screenshot) on your client?

[semion1956]

@matanzuckerman To be completely clear... I mean - when you launch info.py, and webbrowser is called, do you see wrong html page (without styles and js) in your client machine?

[tomerm]

@matanzuckerman screenshot can't be from git. We have not uploaded any screenshots to git. Most probably you downloaded the content of git to your computer and double clicked on the HTML file. This is NOT the way to see it. By the way, HTML file in git is just a sample. I am not even 100 % sure that its' relative location in git corresponds to actual location on local file system when it is getting generated by executing info.py

Please execute, info.py for generation and view of HTML file on your side. If this does not work for you let us discuss why. This is currently the only supported way to formally see HTML file.

[matanzuckerman]

@semion1956 @tomerm

The browser is lunched when I'm running info.py from the server. when I'm running the script from the client GUI the browser it not lunched (probably I need to configure it to run the browser on the client). We are mostly working in client-server environment.

Second, after running info.py from the server and the browser did opened for me I still saw the same landing page (without any results). It is the same like I attached in the screenshot before. The server is lunching Mozilla browser.

[semion1956]

@matanzuckerman Now content of css and js files is embedded into html, so you should not have the problem with relative paths. Please try.

[matanzuckerman]

Hi @semion1956, There is still problem with the html. When running info.py the browser do open (when I'm running from the server) but I see the same as I open it manually. I do get this warning which might be the reason for the problem - "file MLClassificationData/reports/curInfo.html doesn't have json format. Skipped." What can I do?

Thanks

[semion1956]

@matanzuckerman 1) info.py generate html from all json files, which exist in the folder, defined in config. After html is generated, it can be launched by browser, because now it doesn't contain any references to actual files. 2) generated html is placed into the same folder, as json. Message from info.py isn't a problem, but warning - it simple say, that found some file not in json format.

[matanzuckerman]

@semion1956

In this case the html is empty =/ I don't see any results.

[semion1956]

In which case, Matan? Please read README, check your config file. In the worth case - send me your config and html.

[matanzuckerman]

@semion1956 I'm following exactly as it written in the readme file and still I'm getting an html like the photo I uploaded in the first message. Unfortunately It's not simple to send you the files because of the client's privacy.

2. There is a bug with infofrom. when choosing "today" the code is falling because in "luncher.py" you set the parameter "chk" if infofrom != "today" but then when it's equal to "today" "chk" is not defined

[semion1956]

@matanzuckerman Bug with 'today' in infofrom is fixed, thank you. What about "empty" html, I can only suppose, that you use previous version of the code, where js functions and css styles are given from files. Please check, that header of html doesn't contain references to files. BTW, I place into git example of such html. Just as it doesn't have references to files, you can download and launch it directly in the browser.