doi-utils-get-json-metadata should properly decode the data retrieved

[piater]

doi-utils-get-json-metadata contains the following code:

      (with-current-buffer
          (url-retrieve-synchronously
           ;; (concat "http://dx.doi.org/" doi)
       url)
    (setq json-data (buffer-substring url-http-end-of-headers (point-max)))

For me, these cause two problems:

• url-retrieve-synchronously does not decode the buffer but returns the raw bytes. This can cause the retrieved json data to contain garbage.
• The current buffer is used for retrieval. In some use cases, this requires setting up a dedicated buffer beforehand. I suggest making this function robust and self-contained by using its own temp buffer.

Thus, I suggest replacing the above code by the below:

      (with-temp-buffer
        (url-insert-buffer-contents (url-retrieve-synchronously url) url)
    (setq json-data (buffer-string))

This works for me but I cannot be 100% sure that the current→temp change does not break something somewhere...

[jkitchin]

hm. interesting. I use that pattern a lot in that file (and other places too). I guess it is mainly an issue for loading json though?

[piater]

Perhaps so. This is the only place it bit me. As long as you only extract ASCII material it should not matter. I also use your simple pattern in other places, without problems.

Actually, url-insert is sufficient (the extra power of url-insert-buffer-contents should not be needed) and does not require its own url argument.

[jkitchin]

any chance you have a doi where the current method fails?

[jkitchin]

I think the commit above should resolve this.