.. _fwflag_add_timexdiff:
================
--add_timexdiff
================
Switch
======

--add_timexdiff

Description
===========

Creates a feature table of the (mean, std) difference between each message's post-time and the time expressions mentioned within it (e.g. "see you tomorrow" vs. when it was posted).

Argument and Default Value
==========================

No argument. Boolean switch, default: ``False``. Uses :doc:`fwflag_date_field` to know each message's post-time.

Details
=======

Requires a running **Stanford CoreNLP server** on localhost (``FeatureExtractor.addTimexDiffFeatTable()`` connects via ``jsonrpclib.Server("http://localhost:%d" % serverPort)``) with time-expression annotation already available — not something set up in this pass.

**Known issue — broken in SQLite mode too, and hit first.** Verified: under a SQLite corpus, this fails before ever reaching the CoreNLP server logic — ``mm.disableTableKeys(...)`` (same pattern as :doc:`fwflag_add_char_ngrams`) throws ``sqlite3.OperationalError: near "DISABLE": syntax error`` while creating the destination table.

Other Switches
==============

Required Switches:

* :doc:`fwflag_d`, :doc:`fwflag_c`, :doc:`fwflag_t`
* :doc:`fwflag_date_field`
* A running Stanford CoreNLP server (external, not installable via pip)

Example Commands
================

Requires MySQL and a running CoreNLP server — not runnable as-is against the packaged SQLite ``dla_tutorial`` dataset:

.. code-block:: bash

	dlatkInterface.py -d dla_tutorial -t msgs -c user_id --date_field date --add_timexdiff