[WIP] Docker container with a html file server

[eitsupi]

Early steps in a approach to solving #19 with a html file server.

I'm using go-http-file-server on this trial basis now.

1. Upload a pdf file {filename.pdf} to the file server localhost:8080.
   
2. Open the url http://localhost:5000/paper2html?url=http://htmlserver:8080/{filename.pdf}.
   

[ktaaaki]

To omit the step 2 (pasting the request url into the browser), you can automatically convert a pdf triggered by the addition of the pdf to the file server.

The code to convert a pdf from folder monitoring is as follows.

class PdfFileEventHandler(PatternMatchingEventHandler):
    def __init__(self, patterns=("*.pdf",), ignore_patterns=None, ignore_directories=True, case_sensitive=False,
                 debug=False):
        super().__init__(patterns, ignore_patterns, ignore_directories, case_sensitive)
        self.debug = debug

    def on_any_event(self, event):
        if not os.path.exists(event.src_path):
            return

        n_div_paragraph = math.inf
        line_margin_rate = None
        verbose = self.debug
        Paper.n_div_paragraph = n_div_paragraph
        paper2html(event.src_path, cache_dir, line_margin_rate, verbose)

See watchdog_part branch for details.

[eitsupi]

@ktaaaki It looks work!

However, it seems that extra triggers can generate extra files (test-0 and test-1).

[ktaaaki]

That was because the file system watcher is different from my environment.
I fixed this bug in watchdog_part.

[eitsupi]

@ktaaaki I tried the latest watchdog_part d703c58 and It seems that no extra files are generated anymore!

But, I found another bug.
go-http-file-server can't open directories which contain %2F(/) in them names generated by online pdf files.

open ↑ directory.

This is a bug in go-http-file-server, but I think it can cause the same problem in other environments.

[ktaaaki]

Percent encoded slashes may make those errors.
OK, I renamed the name of the directory and changed structure of tree of output directories.