Trendlines

+ Fixed a major issue where dataloss would occur when performing database
migration 0005 -> 0006. (158)
+ Links to the Swagger/ReDoc API documentation are now provided on the main
page. (152)
+ Fixed small error in development documentation. (153)
+ Plots are now all shown on a single page, thus making it much easier to
navigate between plots. (58)
+ Added Bootstrap4
+ updated the jstree data structure to use internal ID instead of a
generated URL (156)

Notes
This revision contains some breaking API changes.

Changes
+ `create_db` was moved out of `app_factory.py` and into `orm.py`. (115)
+ All files created during tests are now made in the `/tmp` directory. (44)
+ Migrations are now performed automatically when the flask app is created
(when the first request comes in to WSGI). (71, 114)
+ Dropped support for Python 3.5 (119)
+ Added the REST API to get all metrics (120)
+ `db.add_data_point` was renamed to `db.insert_datapoint` to match with
future API naming conventions. (123)
+ Added an internal api for datapoints to `db.py`. (125)
+ The REST api is now viewable via a Swagger (or ReDoc!) web page! (34)
+ Core API routes have been refactored to use Flask's Pluggable Views. (128)
+ Sorted the Swagger API into logical blueprints by Schema/MethodView. (133)
+ Replaced all instances of port 9999 with port 2003. (104)
+ The `missing_required_key` error response now accepts lists or tuples.
(137)
+ The `/data` API routes have been moved to a MethodView class. (138)
+ **BREAKING** The api has been refactored to be more consistent. Public API
on core schemas (`Metric`, `DataPoint`) will only accept internal IDs and
no longer accept strings for metric identification. See the docs for
additional details. (140)
+ Minor updates to how the `populated_db` test fixture works (142)
+ Table PKs are now explicit `AUTOINCREMENT` rather than using SQLite's
internal `ROWID`. This will result in increased cpu/memory/disk overhead,
but it ensures that PKs cannot be resued. (143)
+ The REST API for `DataPoint` has been implemented. (122)
+ Example `docker-compose.yml` file now includes `depends_on`. (99)
+ A tip on generating a secure secret has been added to the installation
docs. (93)
+ Command line stuff is now handled by `click`. (150)

Notes
Upgrading from 0.4.0 to 0.5.0 breaks the migration history of the
database. **Automatic upgrades are not supported.** In order to correctly
upgrade without losing your data, follow these instructions (assuming
docker-compose):

> Note: you may need to add `sudo` to most of these.

1. Bring `trendlines` down: `docker-compose down`.
2. Backup your database: `cp internal.db internal.db_old`
3. Pull the new `trendlines` code: `docker-compose pull`
4. Delete your old DB and make a new, empty database file (see issue 110):
bash
$ rm internal.db
$ touch internal.db
$ chown www-data:www-data internal.db

5. Run the migrations on this new file:
bash
$ docker-compose run --rm --no-deps trendlines \
peewee-db \
--directory /trendlines/migrations \
--database sqlite:///data/internal.db \
status
$ docker-compose run --rm --no-deps trendlines \
peewee-db \
--directory /trendlines/migrations \
--database sqlite:///data/internal.db \
upgrade

You should see the following outputs, perhaps with a "Creating network"
thrown in there from docker:

INFO: [ ] 0001_create_table_metric
INFO: [ ] 0002_create_table_datapoint
INFO: [ ] 0003_add_spec_limits
INFO: [ ] 0004_unique_constraint_metric_name
INFO: [ ] 0005_on_delete_cascade_metric

INFO: upgrade: 0001_create_table_metric
INFO: upgrade: 0002_create_table_datapoint
INFO: upgrade: 0003_add_spec_limits
INFO: upgrade: 0004_unique_constraint_metric_name
INFO: upgrade: 0005_on_delete_cascade_metric

6. Copy your data from the old file to the new:

$ sqlite3 internal.db
sqlite> ATTACH 'internal.db_old' as old;
sqlite> INSERT INTO metric SELECT * FROM old.metric;
sqlite> INSERT INTO datapoint SELECT * FROM old.datapoint;
sqlite> .q

7. Bring up `trendlines`: `docker-compose up -d`
8. Verify that you can add a new metric and datapoint:
bash
$ echo "some-new-metric-name 99" | nc $SERVER $PORT


Changes
+ Fixed an issue with migrations (112).

New Features
+ Added columns to the `Metric` table to support limits. (65)
+ Many more API routes have been added:
+ `GET /api/v1/metric/<metric_name>` has been implemented (73)
+ `DELETE /api/v1/metric/<metric_name>` has been implemented (78)
+ `POST /api/v1/metric` has been implemented (74)
+ `PUT /api/v1/metric/<metric_name>` has been implemented (75)
+ `PATCH /api/v1/metric/<metric_name>` has been implemented (83)
+ `DELETE /api/v1/datapoint/<datapoint_id>` has been implemented (57)
+ Implemented database migrations (62)
+ The program version is now displayed on all pages. (109)
+ The tree is now auto-expanded by default. (105)

Changes
+ Changed (again) how we handle being behind a proxy. (60)
+ DB migrations are now inlcuded in the docker image, and documentation
was added on how to perform upgrades. (67)
+ The `Metric.name` column is now forced to be unique. Previously this was
enforced on the software side, but not on the database side. (66)
+ The `DataPoint.metric_id` foreign key is now set to CASCADE on deletes (69)
+ Error responses for the REST API have been refactored (85)
+ Additional tests for PUT/PATCH metric have been added (86)
+ Make use of peewee's `playhouse` extensions for converting model instances
to and from dicts. (87)
+ `peewee-moves` was updated to v2.0.1.
+ Documentation is now reStructuredText and is hosted by ReadTheDocs (91)
+ Switched to using `loguru` for logging. (94)
+ Renamed some function arguments to be more clear. (89)
+ Removed a hack that caused plot urls to be generated client-side. Was
blocking 47. (100)
+ Moved javascript out of HTML files. (47)

+ `units` are now also returned in the GET `/data` API call.
+ Removed a confusing route: `/api/v1/<metric>`. (39)
+ Added a title that will link back to the index page. (40)
+ Changed the way we handle generating links when behind a proxy that's
mucking about with the URLs. (41)
+ Units will now be displayed on the y-axis if they exist. (37)
+ `routes.format_data` was moved to the `utils` module.
+ Added support for optional Celery workers that accept TCP data, allowing
for easier transition from `Graphite`. (6)

+ [BREAKING] Updated REST API: "add" and "get" changed to "data". The HTTP
method will be used to determine adding or retrieving.
+ Implemented switching between sequential and time-series data. (8).
+ Plot link generation has been moved client-side. This fixes an issue where
links would not work when an external force modifies the URL. Fixes 33.