Skip to content

Latest commit

 

History

History
181 lines (142 loc) · 7.35 KB

README.md

File metadata and controls

181 lines (142 loc) · 7.35 KB
Raw

'Plotter' - make a graph

A collection of utilities for reading and modifying EOL graph databases.

Installation

Clone the repository.

Configuration

Copy config/config2.sample.yml to config/config2.yml and modify as follows to make it appropriate for your local plotter installation.

  1. Configure neo4j URL, user name, password, and graphdb name
  2. Set locations: workspace: path: to a local directory where the plotter scripts can put results and intermediate files. Personally I set this to directory .plotter_workspace in my home directory.
  3. Set prod_pub: token_file: to be the local file that contains (or will contain) a token for an EOL v3 API endpoint; similarly for beta_pub. (see note, below)
    1. Obtain a read token for the graphdb, and if you'll be writing as well, also obtain a write token. The write token must be associated with an admin account; a read token can be associated with any account but using a non-admin account is a bit more secure. If using an admin account the read and write tokens will be the same.
    2. To obtain a token, go to https://eol.org/services/authenticate or https://beta.eol.org/services/authenticate as the case may be (see API documentation).
    3. Put the token into a file.
    4. Set prod_pub: token_file: (or beta_pub: token_file: or prod_pub: update_token_file: etc.) to be the path to that file.
  4. Configure the staging server locations: staging:. The directory on the staging server has to be writable via the rsync command, which is used by plotter scripts, and it has to be readable by the Neo4j server(s) via HTTP. The examples in config2.sample.yml should provide guidance on how to set these variables.
    1. rsync_command specifies the rsync-like command to use to transfer local files to the staging server (this string does not include the source or target).
      If not specified, defaults to rsync -av.
    2. rsync_specifier should be either a local path or host:path as would be understood by an ssh command. host is either a DNS name or a name configured in ~/.ssh/config. If a local path, then neo4j should be running locally and the path should be the path to neo4j's import directory. If remote, the directory should be one that is exposed by that host's HTTP server.
    3. (It may be possible to use the neo4j import directory with file:/// URLs, but I haven't tried it.)
    4. url gives the prefix designating the rsync_specifier directory when accessed using HTTP. It will be used in the URLs that will occur in neo4j LOAD CSV commands.
    5. In order for rsync to work, appropriate ssh credentials have to be in place. They can be specified in ~/.ssh/config or with an -I argument in the rsync_command.

Choosing EOL servers

Most of the rake commands require a CONF= parameter to specify which publishing and/or graphdb server is to be contacted (and, in turn, which EOL repository (content) server is to be consulted). The graphdb choices are listed in config2.yml but are typically test (for a private testing instance), beta (EOL beta instance), or prod (EOL production instance).

'Smoke test'

As a simple first test of installation, try

rake resource:info CONF=beta ID=40

which will display a bunch of information about resource 40 on the beta publishing instance.

All-traits dump

Plotter can generate a zip file with a dump of all trait records. See doc/trait-bank-dumps.md.

Trait inference ("branch painting")

Plotter can do "branch painting" or inference of traits through the taxonomic hierarchy. See doc/branch-painting.md.

Resource metadata

There is a script for copying resource metadata into the graphdb. For example:

rake traits:sync_resource_metadata CONF=prod

ensures that there is a neo4j Resource node for every resource known to the production publishing server, and sets the resource_id, name, description, and repository_id property of each.

Normally it won't be necessary to invoke this script since this information is also transferred on a 'publish'.

Dynamic hierarchy

There are scripts for adding ranks (see rake --tasks hierarchy:sync_metadata) and vernacular names (see rake --tasks vernaculars) to the graphdb. It is also possible load a dynamic hierarchy into a test instance (see rake --tasks hierarchy).

Deltas

There are some general scripts for manipulating CSV files, and in particular for comparing them ('delta' or 'diff'). See doc/csv.md.

Retrieve DwCA for a resource

See rake --tasks resource.

Workspace structure

Workspace root comes from config2.yml (via system.rb). Default is /home/jar/.plotter_workspace, which should be changed.

Sometimes the choice of location is a bit arbitrary or even wrong. In general artifacts that depend only on the repository goes in xxx_repo (for xxx = prod, beta, test); if it depends on the additionally on the publishing relational database it goes in xxx_pub; and if it depends additionally on the graphdb it goes in xxx.

(workspace root)/ dwca/ NNN/ ... one directory per DwCA ... id = final 8 chars of uuid properties.json ... metadata for this DwCA dwca.zip or dwca.tgz unpacked: meta.xml (all the other .tsv or .csv files) prod/ resources.csv prod_pub/ resources/ ID/ ... one per publishing resource id inferences/ - temporary directory for intermediate files inferences.csv prod_repo/ resources/ REPO_ID/ ... one per repository resource id page_id_map.csv pages/ accepted.csv.chunks/ beta/ beta_repo/ ... same structure as or production ...

An 'instance' is a triple (graphdb, publishing instance, repository instance).

Testing

Lots of things to test. For end to end tests we need to look at:

  • painter - branch painting
  • traits_dumper - copy traits from graphdb to a set of files
  • traits_loader - inverse of traits_dumper
  • resource - copy vernaculars into graphdb
  • cypher - run a single cypher query
  • hierarchy - dynamic hierarchy load, diff, and patch
  • instance - flush caches (this has fallen into disrepair)

The 'concordance' feature is not currently working.

Upgrading from plotter 0.1.0 to plotter 0.2.0

  • Make sure you have a version of the bundle command with version >= 2.2.10. I did gem install bundler followed by bundle to accomplish this.
  • Make a new configuration file with cp config/config2.sample.yml config/config2.yml.
  • Add details to config/config2.yml based on what you did for your previous config file config/config.yml. The structure and syntax are slightly different but overall the details are mostly the same. I hope that this step is self-evident.
  • Traits dumps now go in workspace subdirectory prod/trait_dumps (for production). Modify scripts for this new location as necessary.
  • Traits dump temp files will now go in workspace subdirectory prod/trait_dumps/tmp/
  • Be sure to include R among the rsync flags (see rsync_command:)