UPDATE on 12-May-2018: You can now directly test my work by cloning my repo, updating the source tree to my bookmark by hg up ocs, making a build and calling wiki_login from octave-cli to get a login token in return. I’ve added the files and necessary changes in the codebase itself. There’s no directory ocode now.

If you happen to already have a build of octave, just do the following:

  • cd path/to/your/source/tree
  • hg pull https://me_ydv_5@bitbucket.org/me_ydv_5/octave in your source tree.
  • hg up ocs
  • hg up -r 8bbf393
  • make -jX in your build tree.

This will save your time of cloning the entire repo and compiliing.

As mentioned in the previous post, I worked on __publish_wiki_output__.m and publish.m code. The __publish_wiki_output__.m has been added as an internal function in scripts/miscellaneous/private. I skimmed through the parser in publish.m to get a gist of how it actually works. It has three levels of parsing:

  1. Extract the overall structure (paragraphs and code sections).

  2. Parsing the content of a paragraph.

  3. Generate the output of the script code and look after figures produced in code.

After that, I studied how the url-transfer.h file is implemented which contains a base class named base_url_transfer which has a derived class named url_transder. One thing that puzzled me while doing so was, why there has to be a macro HAVE_CURL in order for curl_transfer to be defined and why we haven’t defined url_transfer class itself? I would try to get these doubts solved this week.

The problem of user agent was solved by selecting the following user agent:

    GNU Octave/OCTAVE_VERSION (https://www.gnu.org/software/octave/ ; help@octave.org) libcurl/LIBCURL_VERSION

where OCTAVE_VERSION and LIBCURL_VERSION correspond to the user’s octave and libcurl version respectively. This code precisely does the same for us.

My inteded plan for the wrapper is, make a cookie_manager.m file that will process the various user options (like verbose output, timeout settings, api.php url, etc.) and pass the values to an internal __curl__.cc function which will in turn, take help from libcurl_wrapper.cc to do various tasks (all the work related to cookies will be looked after by it, essentially).

Curently, all the code in wiki_login.m has been commented out except the first step of login, i.e, getting a login token from the api.php, which it smoothly does, as of now. I am assuming that the file which would store the cookies, is temporary and should be deleted once the session expires. This is one of the things I will be looking on in this week.

I’ve migrated all the developments from my forked git repo to my mercurial bookmark ocs recently and so I was not sure where should I put the files in my source tree. Thus, I’ve put all of them in a directory ocode for now.

To test this for yourself:

  • Clone my build tree using hg clone https://me_ydv_5@bitbucket.org/me_ydv_5/octave
  • Make yourself a build of octave (make -j2, etc.).
  • cd octave.
  • Update to my bookmark using hg up ocs. (IMPORTANT!)
  • cd ocode.
  • Execute Makefile in octave.
  • Execute wiki_login in octave to get a login token.

All other details of the wrapper’s implementation will be followed in the next post.

The next week would follow the following advances:

  • Choosing the right location for the files (after I get a green light for the current developmental path).
  • Extending other options in the wrapper for wiki_login’s steps 2 and 3.
  • Implementing cookie_manager with other user options.
  • Writing help text and text cases, if any.
  • Correction of existing work/ changing the stategy as adviced by mentor, or anyone else.
  • Look into how can I use existing base_url_transfer class in the wrapper and resolve my query of the HAVE_CURL macro and shared pointers, etc.

I am optimistic that I would be able to complete my first evaluation work by 25th May or so, as I will need to focus on my end term examinations after that which will start from 1 June. We don’t get holidays in between the exams!

Please let me know if I am doing it the right way or not, by either replying to this thread or by simply dropping a message on #octave for <batterylow>. All the suggestions are always welcomed.

Oh and not to forget, I got my domain an SSL certificate, now all the requests are served via HTTPS only!

Stay tuned for next update.