* Refactored ``ts_json:ext_data_to_ejson{1,2}`` from a function that iterates
over a list of extended data properties and returns a transformed list to
``ts_json:ext_data_to_ejson/1``, a function that transforms one extended data
property. The iteration behavior is acheived using ``lists:map/2``.
* Refactored ``ts_json:ejson_to_ext_data{1,2}`` in a similar fashion to
``ts_json:ext_data_to_ejson{1,2}``.
* Fixed ``ts_json:construct_record/3`` to use ``ts_json:ejson_to_ext_data/1``.
------------------
* Created the extended data table. This is a more generic version of the
extended data field that was on ``ts_user``. Instead of arbitrary key-value
pairs going in a list on the user record we will have an additional table, a
one to many relationship between existing tables and the ``ts_ext_data``
table, each row being an extended value of the corresponding record.
* Added support to the ``timestamper:create_tables/1`` and
``timestamper_dev:create_table/1`` functions for the new ``ts_ext_data`` table.
Documentation
-------------
* Added some general docs about the DB layer code.
* Added two new issues, *D0020*: Entry exclusion filters and *D0021*: notes width.
API Changes
-----------
Necessitated by the change to the data model.
* Updated ``ts_api`` data-retrieval functions to use extended data:
* ``get_user_summary/2``
* ``list_timelines/2``
* ``list_entries/3``
* Updated ``ts_api:put_timeline/3`` to parse the extended data supplied by the
caller. *FIXME* it is not actually saving this data.
* TODO: Started updating ``ts_api:post_entry/3`` to handle extended data, need
to finish.
Database Layer
--------------
Changes necessitated by the change to the model.
* Added ``ts_common:do_set_ext_data/2`` which iterates through the extended data
key-value pairs calling ``ts_ext_data:set_property/3``. It does not provide a
transaction context.
* Split ``ts_common:new/1`` and ``ts_common:update/1`` functions into multiple
functions to prevent code-duplication:
* ``do_{new,update}`` contains the code that performs integrity checks and
the actual database write function. It does this assuming that it is being
called from within the context of an mnesia transaction (uses
``mnesia:read`` and ``mnesia:write``).
* ``{new,update}/1`` performs the same function as previously. The
implementation changed from using mnesia ``dirty_*`` calls to prociding a
transaction and calling ``do_{new,update}/1``.
* ``{new,update}/2`` expect the record to update/create and the extended
data to write atomically with the record. They provide a transaction
context, call ``do_{new,update}/1``, then call ``do_set_ext_data/2``.
* Similar to the refactoring of ``ts_common:{new,update}/1``,
``ts_entry:{new,update}/1`` have been refactored into multiple methods each to
support extended data properties:
* ``do_{new,update}/1`` perform the actual update assuming we have already
established an mnesia transaction.
* ``{new,update}/1`` behave the same as they used to, but now do so by
creating an mnesia transaction and calling ``do_{new,update}/1``.
* ``{new,update}/2`` create an mnesia transaction, call
``do_{new,update}/1``, and then call ``ts_common:do_set_ext_data/2``.
* Again similar to the refactoring of ``ts_common:{new,update}/1``,
``ts_user:{new,update}/1`` have been refactored into multiple methods each to
support extended data properties:
* ``do_{new,update}/1`` perform the actual update assuming we have already
established an mnesia transaction.
* ``{new,update}/1`` behave the same as they used to, but now do so by
creating an mnesia transaction and calling ``do_{new,update}/1``.
* ``{new,update}/2`` create an mnesia transaction, call
``do_{new,update}/1``, and then call ``ts_common:do_set_ext_data/2``.
* Created the ``ts_ext_data`` module as the interface to the extended data
properties introduced in the data model:
* ``create_table/1`` performs the same function as it does in the other db
layer modules, creates the table with the appropriate structure given the
more general table options desired (location, storage type, etc.).
* ``set_property/3`` takes a record, a property key, and a property value as
input and sets the property described by the property key on the record to
the given value, assuming this is a valid property for the record to have.
For example, currently the ``ts_user`` record can have an associated
property, ``last_timeline``, which represents the last timeline the user
was working with. Trying to pass this property with a ``ts_entry`` record
would result in an exception. This function uses
``ts_ext_data:do_set_property/3`` as its underlying implementation.
* ``get_property/2`` takes a record and a property key and returns the value
of that property for the given record, or ``not_set`` if the property has
not been set on that record. This method creates its own mnesia
transaction.
* ``get_properties/1`` takes a record and returns a list of key-value tuples
representing all of the extended data properties set for the given record.
This method creates its own mnesia transaction.
* ``do_set_property/3`` takes a record reference (not the whole record), a
a property key, and adds the property assignment to the ``ts_ext_data``
table. It creates its own mnesia transaction.
* Added ``new/2`` and ``update/2`` to the ``ts_timeline`` module to support
extended data properties. They delegate implementation to
``ts_common:{new,update}/2``.
JSON Encoding/Decoding
----------------------
Changes necessitated by the change to the data model.
The JSON objects now contain a potentially unlimited number of fields, as each
extended data property is encoded as a seperate field, and looks no different
from any of the required fields on the object. The intended explanation in API
documentation is that each object type (``user``, ``timeline``, or ``entry``)
now has both *required* fields that *MUST* be present in every message in either
direction and *optional* fields that may or may not be present in any
communication with the API. There should be a clear distinction between which
fields are required and which are optional. It might also be a good idea to
provide a suggested default for optional values when they are not present.
* Updated documentation about JSON record structures to reflect the fact that
there are now potentially many optional attributes in addition to the required
attributes for each record.
* ``record_to_ejson/1`` refactored to ``record_to_ejson/2`` which also takes
the extended data attributes and appends them as additional attributes to
the end of the record structure.
* Created ``ext_data_to_ejson/{1,2}`` to provide a mechanism for reformatting
extended data properties whose internal representations are not immediately
translatable into JSON. Currently only the ``entry_exclusions`` property,
which is a list of strings, needs to be treated this way (changing from
``[val, val]`` to ``{array, [val, val]}`` as needed by ``json:encode/1``.
``ext_data_to_ejson/1`` acts as a more user-friendly facade to
``ext_data_to_ejson/2``.
* Rewrote ``ejson_to_record/{2,3}`` and ``ejson_to_record_strict/{2,3}`` to
handle extended data. They now use a common method, ``construct_record/3`` to
create the actual record object and extended data key-value list.
``ejson_to_record_strict/{2,3}`` only differs in that it checks for the
presence of each required field of the record after the record is constructed.
The three-parameter versions of these functions also take in the intended
reference for the constructed record, replacing anything that is in the EJSON
body as the record reference (useful when the body does not have the record
ids). These methods now return a tuple: ``{Record, ExtData}`` instead of just
the record.
* Created ``construct_record/3`` takes a record and the EJSON fields from the
input object. The third parameter is an accumulator for the extended data
properties found when constructing the record. This method works by iterating
over the list of input fields. It recognizes any required fields and updates
the record being built with the value. Any fields it does not recognize it
assumes are extended data properties and adds to its list. When all input
fields have been visited it returns the record and list it has constructed.
* ``ejson_to_ext_data/{1,2}`` is the inverse of ``ext_data_to_ejson/{1,2}``.
*TODO*: this method is not actually being used by the ``ejson_to_record*``
methods.
* Resolved issues:
* #0006: Fix timeline menu UI.
* #0015: Create new timeline button in timeline menu.
* #0017: Implement timeline creation.
* Removed ts_api:post_timeline/3, corresponded to `POST` to
`/ts_api/timelines/<user-id>`, which makes no sense as there is no way to
auto-generate new timeline-ids.
* Changed ts_api:put_timeline/3 to use ts_timeline:write. This resolved#0017.
* Changed ts_api:put_entry/4 to use ts_entry:write.
* Implemented ts_entry:write/1 and ts_timeline:write/1. These functions are
simple passthroughs to mnesia:dirty_write/1.
* Added clarifing documentation for ts_json:record_to_ejson JSON formats.
* Renamed ts_json:ejson_to_record/2 to ts_json:ejson_to_record_strict/2 and
added ts_json:ejson_to_record/2 as a lax version of the same.
* Updated all ts_api functions to use updated ts_json methods.
Using atoms will not scale in the large. Also, using atoms as keys forced the
API to convert arbitrary end-user input to atoms, adding another potential
drain of the finite atom-space available.
- ts_api: url paths are now treated and matched as lists, not atoms.
Usernames and API calls all use lists now.
- ts_json: key items (username, timeline ids) are now expected to be lists,
not atoms.
- Instead of returning Meta-Data (status) in the header and the body, we now
only return the content in the body, except in error when we return the error
message.
- JSON entries use the 'id' name for entity ids (instead of 'username',
'timeline_id', and 'entry_id')
- Bug fix in ts_api:list_entries/3. Case statement matching on atoms but
input is a list (string).
- Bug fix in ts_api:put_entry/3. Was expecting the wrong result from
ts_entry:new/1.
- Bug fix in ts_entry:list/4. Code crashed when the starting offset was
greater than the total number of elements. Now returns [].
- Fixed ts_json:encode_datetime/1 and ts_json:decode_datetime/1 to handle
millisecond values in the datetime string (per ISO standard).
- Broke out ``control-links`` style to a top-level class.
- Added showdown.js, a JS Markdown processor. Not hooked up to anything
yet but intend to display entry notes with Markdown.
- Added code for entry pagination. Loads the most recent 20 entries and
loads more upon demand in batches of 20.
- Fixed bug in login routine that kept the user edit fields from being
pre-populated.
- Rewrote the loadEntries function to double for new entries and loading
more existing entries.
- Commented displayEntries. Also refactored into displayNewerEntries,
which pushed new entries on to the top of the stack, and
displayOlderEntries, which tags them onto the bottom.
- Implemented hidden notes field for new entry input.
- Implemented new entry creation.
- Created a helper function to ISO format a Date object.
- Expanded entry template to show control links (edit, show notes, del).
- Activated the 'load more entries' button.
- Bug fix in ts_entry:new/1. Msspelled ``atomic``.
- Bug fix in ts_json:record_to_ejson/1. For ``ts_entry`` records, the
Username and TimelineId elements were not being converted from atoms to list.
- Added the entry template for loaded and created entry elements.
- Added ICanHaz.js (which wraps mustache.js) and underscore.js.
- Implemented a naive version of displayEntries() in ts.js.
- Added debug alerts for error cases in ts.js.
- Styling the new entry elements.
Created timestamper module to start the application.
Added cookie-based authentication to ts_api.
Added utility methods to ts_api:
* make_json_400/1 and make_json_400/1
* make_json_401/1 and make_json_401/2
* parse_json_body/1 reads a JSON object from a HTTP request body.
Implemented ts_api_session module to manage api user sessions.
Fixed ts_entry:list* methods to be 0-indexed.
Removed the ts_json:ejson_to_record/1 implementation for ts_user records.
Decided that ts_user records are never trusted from the client,
manipulation of fields such as pwd, username will be restricted to
app pages.
Changed the password hashing algorithm. Now uses SHA1(pwd + 256bit salt).
Want to use bcrypt, investingating cross-platform bcrypt implementation.
Fixed yaws.conf config file.
Implemented ts_api:list_timelines/2.
Adjusted ts_timeline:list/3 to be 0-indexed.
Changed ts_user password hash to use a random salt + SHA1
Added some skeleton testing code.
Created a utility function for OK results, ts_api:make_json_200/2
Created ts_common:new/1 and ts_common:update/1 to generalize record creation and update.
Refactored ts_timeline:new/1 and ts_timeline:update/1 to use the ts_common functions.
Implemented ts_json:record_to_ejson/1 and ts_json:ejson_to_record/1 for ts_user records.
Implemented ts_user module.
The id_counter module now includes the record directly in the source.
Fixed many typos and small syntax errors.
Added ts_api:dispatch_user/2 to handle different HTTP methods.
Added method placeholder stubs to ts_api.
Implemented ts_api:list_entries/3.
Added ts_user record and ts_user module.
Implemented ts_entry:list/4, the more generic guts of the other list functions.
Created ts_entry:list_asc/3 and ts_entry:list_desc/3.
Fixed ts_json:encode_datetime/1.
