blastochor 0.1.2

Harvest, transform, and export data from Te Papa's collections API.

Blastochor: Harvest data from Te Papa's collections API

"blastochory, where the stem of the plant crawls along the ground to deposit its seed far from the base of the plant"

This tool lets you access Te Papa's collections API using some basic configuration options, transform the data based on a mapping file, and export the data as a single CSV or multiple CSVs that reference one another, such as can be combined in a Darwin Core Archive (DwC-A).

This repo includes a basic mapping file that will be used if another is not provided.

API documentation

Still to do

• A complete default mapping that dumps out as much data as possible
• JSON export
• Logging functionality
• Command line options to run a search or single record, returning json or flattened default csv
• New kind of output that relates records together based on given criteria

Installation

Clone this repo using git clone. Install the following packages:

• askCO
• PyYAML
• click

Go to https://data.tepapa.govt.nz/docs/register.html and register for an API key. Add the API key to an environment variable called 'TE-PAPA-KEY'.

Usage

The config.yaml file at the root level sets search and export parameters – this file has useful defaults, so if you want to change it much, create a new one in the projects directory. More on this below.

The main functional setting in the config file is mode: search, scroll or list. You can either send a search query, which will page through and return all relevant records (limit 50K), scroll through all records (no limit) or provide the app with a list of IRNs (internal reference numbers, Te Papa's record identifiers), which will be individually queried and returned.

The default mapping file (resources/mapfiles/defaultmap.yaml) will provide some basic object information. You can create a new one for your project.

When your config and mapping are ready, run the app with python -m app.

When run, the application runs validation on your config and mapping files. If any errors are found the application will exit and you'll find a report textfile at root containing the validation log. If the config's quiet is set to false, it will also be printed to the CLI.

Command line parameters

Blastochor can be set up with projects using a pre-defined config and mapping file. Put the config file in projects with the name [projectname]_config.yaml and the mapfile (set the name in the config file) in resources/mapfiles.

To run a project, enter python -m app --project=[projectname].

You can also set a limit on how many records to retrieve, outside what's in your config file - great for testing. Enter python -m app --limit=[number]. If search or scroll are set up to get more records (eg you ask for 100 but scroll is set up to get 1000 at a time), you'll get the larger number.

Run the application with a profiler using python -m app --profiler=True. A .prof file will be written to the logging directory.

Use multiple parameters by stringing them together: python -m app --project=plants --limit=10000.

Config parameters

Setup

api_key_env: Environment name for your unique API key. Default is TE-PAPA-KEY

mapping_file: Filename for file containing data mapping rules. Default is default.yaml (points to a default mapping inside the app)

input_dir: Directory holding any source files, like a skiplist or list or IRNs. Default is input_files

output_dir: Directory for exported CSVs. Default is output_files

use_skiplist: Set to true if using a skiplist. Default is false

skipfile: Filename for skiplist. Default is null

min_image_size: Only export object records that have images that are at least this many px on each side. Default is 0

max_list_size: Only collate lists of values into a single value up to the specified size. Default is 100

clean_newlines: Removes newlines (\n) from values, as CSVs treat them as literal linebreaks. Default is true.

group_rows: If exploding records into multiple rows, set to true to group them under a parent row. Default is null

parent_label: If grouping rows, sets a label for the parent. Only used if group_rows is true

parent_fields, child_fields, ungrouped_fields: Specify fields that need to be included or excluded for each kind of row. Set include to true to only include the specified fields, or false to include everything except those specified. Only used if group_rows is true

Run

mode: Set to search, scroll, or list. Default is search

quiet: Set to false if you want detailed progress messages written to the CLI. Some broader messages and progress bars will be displayed either way. Default is true

base_url: URL used to query the API. Default is https://data.tepapa.govt.nz/collection/

endpoint: Set to the primary endpoint you want data from, or set to null if you want to search across all endpoints. Defaults to object when querying individual records if not set

timeout: How long in seconds to allow each query before timing out and retrying. Default is 5

attempts: How many times to retry each query before returning None. Default is 3

List mode settings

list_source: Filename for list of IRNs to query

Search/scroll mode settings

record_limit: Set a limit on the number of records you want to export. Default is null (no limit)

size: Set a limit on the number of records returned in one page of results. Set to a maximum of 1000 for scroll. Default is 100

query: String to search for. Default is * (wildcard)

sort: Fieldname to sort results by, such as id, _meta.modified. Add a - before the field to sort in reverse order. Default is null (sorts by id)

Filters:

collection: Set collection to constrain search. Uses the APIs collection labels (PacificCultures, not Pacific Cultures). Default is null

allows_download: Set to true to only return object records that include downloadable images, false to only return records that don't. Default is null

keyword_fields: Fieldnames for extra filters on your search. Separate with |

keyword_values: Values for extra filters. Ensure you have the same number of terms as keyword_fields and they're in the right order. Separate with |

Mapping

Mapping files are YAML (.yaml) files that name the CSVs that should be created, triggers for harvesting additional records, and lines of rules with the export-side fields in each file. Each rule then includes the API source field used and the transformation rules applied.

label is what the app uses to generate output files' names and ensure it's applying the right rules to the right records.

reference_column adds a column to the start of the CSV using the provided fieldname, providing a place for non-core files to reference the core file in each row.

primary_endpoint is the endpoint (eg object, agent, taxon) this file will mainly use. Other endpoints can be specified in certain processing functions.

explode sets a field containing list value where each child is written out separately, such as when you want a new row for every image attached to an object record.

reduce cuts down the record's data to just the specified section, for example the list of images under hasRepresentation. This makes it easier to then navigate to subfields. The script expects the reduced data to be a list.

requires checks each record, only letting the record be written if the specified field matches a specified value. If exploding the record, the check will run for each row. If reducing the record, make sure the specified field path takes account of that. For example:

reduce: hasRepresentation
requires:
  rights.allowsDownload: true

Extend

extend contains a list of triggers, identifying certain fields that contain IRNs that should also be harvested. For example:

endpoint: fieldcollection
path: evidenceFor.atEvent.id
for_label: event

This checks harvested records for an IRN in the evidenceFor.atEvent.id field. If found, it will send a separate query to the fieldcollection endpoint, storing the record for later use in the event file.

If the record has already been harvested, the record object will just be updated with a flag that it needs to be included in that file.

path can contain multiple values for the same endpoint. Separate these with , .

To output another CSV with the same endpoint as the core file and using the same path, just add it to the same item's for_label with a , . For example:

endpoint: object
path: id
for_label: identification, multimedia

If for_label is null, the record will be saved but not written out anywhere - do this if you just need some of the record's data for another function.

Extension records are batched and retrieved in lots of 250 per endpoint.

Fields

The fields section contains a list of data transformation rules, headed up with the output fieldnames. These need to be in the order you want them to show in your CSV. The rules are applied to every record included in the file.

Each field then contains a list of functions (one or multiple that run in sequence), which contain a dictionary of the API fields, strings, integers, or other parameters they'll be working on.

For example:

- occurrenceID:
  - literal:
      path: pid

This means the file will have a column called occurrenceID, which take a straight copy of the contents of the record's pid field.

When using a field name as a parameter, you need to provide the full path (unless you've reduced the record - see above). If one of the fields in the path contains a list, include an i to show it needs to be iterated through. For example: path: identification.i.identifiedBy.title

Functions can be chained together by adding them as further list items under the fieldname.

Available functions

literal

Copies the value at the given location. If the path includes a list, you can add an integer to copy a specified list member - if no integer is provided for a list, it will just get the first (0) value.

• parameters: path, ordinal
• example: path: pid
• example: path: additionalType, ordinal: 0
• example: path: production.i.creator.title, ordinal: 1

hardcoded

Sets the value to the string provided. Set the parameter to value: explode_ordinal if you want the value to be the row's order within the record.

• parameters: value
• example: value: https://data.tepapa.govt.nz

collate_list

Gets all values of a field within a list, or multiple lists. Automatically concatenated with " | " before writing out. Maximum length of list is controlled in Config.yml.

• parameters: path
• example: path: hasRepresentation.i.rights.title
• example: path: depicts.i.title, influencedBy.i.title

clean_html

Removes unwanted html markup from a value previously returned, such as a description.

• parameters: null

concatenate

Joins values in a list using " | ", after removing None values. Use after calling another function that produces a list.

• parameters: null

conditional

Returns a value for a given field if another field matches a specified value. Both paths need to be in the same section of the record. If the path given points to a list, each child of the list will be checked. If nothing is found with the initial match condition, an optional fallback condition can be tried.

• parameters: path, condition_path, match, fallback_match
• example: path: related.i.contentUrl, condition_path: related.i.title, match: ORCID, fallback_match: Wikidata

country_code

Finds a country name in a record and looks up its ISO 2-character country code. The file containing country codes is at resources/countrycodes.json.

• parameters: path
• example: path: evidenceFor.atEvent.atLocation.country

create_filename

Use after a function that returns a string. Replaces unsafe characters ( , ?, \, :, ; and so on) with an underscore, or removes them.

• parameters: suffix
• example: suffix: jpg

fallback

Use to substitute another value if the original request returns None. Nest another set of functions and parameters underneath.

- format_string:
    string: {} ({})
    path: production.i.contributor.title, production.i.role
    required: true
- fallback:
  - literal:
      path: production.i.contributor.title

This returns a formatted string if both a contributor's title and their role are present. Because format_string uses the required parameter, if there's no role the function will return None, and then just the title will be returned as a literal value.

first_match

Try multiple paths in order and return the value of the first available one. Useful when trying to return the most precise available location.

• parameters: path (separated by , )
• example: path: evidenceFor.atEvent.atLocation.locality, evidenceFor.atEvent.atLocation.stateProvince, evidenceFor.atEvent.atLocation.country

for_each

Use after another function that returns a list. Goes through and performs further transformations on each member. You can add the concatenate function below for_each, or it will automatically happen when writing out. Format this in the mapping document by nesting the rest of the functions below, for example:

- recordedByID:
  - collate_list:
      path: evidenceFor.atEvent.recordedBy.i.id
  - for_each:
    - lookup:
        endpoint: agent
    - conditional:
        path: related.i.contentUrl
        condition_path: related.i.title
        match: ORCID
  - concatenate:
      null

This finds the agent id for every person at a specimen collection event, looks them up one at a time, and checks if they have an ORCID identifier. None values get stripped out once all child functions have run.

format_string

Finds a value and inserts it into a provided string at a specified place. Mark the location for each value using {}. Include an optional third parameter required to return None if any of the values are unavailable.

• parameters: string, path (if more than one, separated with , ), required
• example: string: https://collections.tepapa.govt.nz/object/{}, path: id
• example: string: {} of {}, path: typeStatus, qualifiedName, required: true

lookup

Use after finding an IRN to return the associated record for the next function. Record needs to have been harvested either directly or using extend functionality.

• parameters: endpoint
• example: endpoint: agent

map_value

Use after another function that returns a value – this can be a single value or a list. Each parameter for this function contains a source value and what to map it to. Each value in a list will be checked. If the value doesn't match any parameter's key, it will return None. Note that this function's parameters are dictionary key/value pairs.

• parameters: key/value pairs
• example:

- literal:
  - path: restrictLocalityData
- map_value:
    true: precise location information not given for threatened species

measure_geodistance

Use after another function that returns a fieldcollection record. Finds the distance between the coordinate centroid and the first of a pair of coordinates, giving the uncertainty radius.

• parameters: units - optional, use "m" to return metres, otherwise will return kilometers

must_match

Use after another function that returns a value or list of values. Checks each term against an authority list and only keeps the ones you want. Not case-sensitive.

• parameters: terms (separated with , )
• example: terms: canvas, paper, plaster, cardboard, ceramic, wood, clay

related

Make a fresh query to the special /related endpoint for the current record, returning records that are connected in some way. Can only be used on a complete record as it requires the href value. Set a size to limit the number of results (default is 100), and specify types of records to filter down.

• parameters: size, types (Capitalise and separate with , - note the lack of a space character)
• example: size: 50 and types: Object,Specimen

use_config

Pull a specified term out of the config file. Can also use hardcoded but lets you avoid doubling up if you change the parameter.

• parameters: key
• example: key: base_url

use_group_labels

If grouping rows for each record, lets you set a label for the parent row, each child row, and standalone rows. Values can be null.

• parameters: parent, child, other
• example: parent: sequence, child: image, other: image