6 Inserting New Traits Via the API

This page provides a general description of how to insert trait data via the v1 version of the BETYdb API. For information about accessing data via the v1 BETYdb API, visit https://pecan.gitbooks.io/betydb-data-access/content/API/beta_API.html. For a list of URLs of API endpoints, visit https://www.betydb.org/api/docs.

6.1 Trait Insertion Endpoint.

The path to use for trait insertion is /api/v1/traits(.EXT) where EXT is csv, xml, or json. These are used to submit data in CSV, XML, and JSON format, respectively. If no extension is given, JSON format is assumed. A user must have Creator status (page access level 3) in order to use the trait insertion API.

6.2 Format of Data Files

To see some valid sample files in all three of the supported formats, see Running the Examples below.

6.2.1 Schema for XML Data Files

XML data files must validate against the schema specified by app/lib/api/validation/TraitData.xsd. It is possible to validate files on the command line with xmllint:

xmllint --schema path/to/TraitData.xsd --noout path/to/xml-data-file

Note that data files of other types (CSV and JSON) are converted to XML internally and then validated using this schema.

6.2.1.1 General Outline and Semantics of XML Data Files

• The root level elements is named trait-data-set.

• trait elements appear below the root, either directly, or nested within intervening trait-group elements.

• A trait element must include a mean attribute with a value of type double. An exponent may be used; for example, in place of "0.0123" one may write "1.23E-2".

• A trait element may have the following attributes:

  • An attribute utc_datetime whose value has the form "YYYY-MM-DDZ" or "YYYY-MM-DDTHH:MM:SSZ" which represents the date or date and time the trait measurement was taken. The trailing "Z" emphasizes that the value is in UTC time and is required. If a time of day is given, the symbol "T" must separate the time from the date. If a time is given, fractional seconds may be included.

  • A local_datetime attribute may be used in place of utc_datetime. This represents the date or date and time of a trait measurement in local time. This attribute may only be used if the trait is associated with a site having a specified time zone. The value format is the same as for utc_datetime except there is no trailing "Z".

  • An attribute access_level must be supplied unless it is supplied using the defaulting mechanism (see below).

• A trait element may have the following child elements:

  • site: This may have any of the attributes id, city, state, country, and sitename. It must have enough of these attributes to uniquely identify an existing site. Withing the database table sites, the values of sitename should be unique and sitename is the preferred attribute to use in identifying a site. Unfortunately, uniqueness is not currently enforced and there are in fact several cases of multiple sites sharing the same site name. In general, using the id attribute to identify a particular trait association is strongly discouraged and should be used only when necessary.

  • species: Allowed attributes: id, genus, species, scientificname, commonname, AcceptedSymbol. scientificname is the preferred attribute for identifying a species. Within the database table species, non-blank values of scientificname should be unique but this constraint is not yet enforced. (scientificname is left blank in cases where a species row represents a category of plant; in this case, the commonname column is used to describe the category.)

    If a particular cultivar of the species is intended, a child cultivar element should be included. This element may use either a name attribute (preferred) or an id attribute to identify the cultivar. Cultivar names are guaranteed to be unique within a given species.

  • citation: Allowed attributes: id, author, year, title, doi. The preferred method of selecting a citation is by doi or by author, year, and title (often just author and year will suffice).

  • treatment: Allowed attributes: id, name, control. The preferred method of selecting a method is by name. [In the process of implementation: A citation is required, either directly on the trait or as a default for a group of traits, if a treatment is to be specified. Moreover, the specified treatment must be associated with the specified citation. This will often make it possible to use the name attribute to specify a treatment, since only treatments associated with the given citation will be considered when selecting by name.]

  • variable: This specifies what the trait measure. This element must be included if it is not specified in a defaults element (see below). Allowed attributes: id, name, description. name is the preferred attribute to use to specify the variable and should be unique, but this isn't yet enforced and there are a few cases of duplicates.

  • method: Allowed attributes: name. This element must have a citation child element. (This citation has no ostensive relation to the citation associated with the trait.) Together, the name and the citation should uniquely determine which method is being used. [To do: Constrain the methods table to ensure that this is always possible.]

  • covariates: This element specifies what covariates are associated with a trait measurement. It allows no attributes but must contain one or more covariate child elements. Each covariate element must contain a variable element (specifying what the covariate measures) and have a level attribute (specifying the value of that measurement).

  • entity: Allowed attributes: name and notes. An entity with the given value for name and notes will be created if no entity with the given name exists. [To be implemented: It is an error to specify an entity at the trait level having a blank name. It is an error to supply a notes attribute if name refers to an existing entity.] [To do: Guarantee uniqueness of non-blank names in the entities table.]

  The eight elements just mentioned specify how the trait is associated with data in other tables. In addition, a trait may include two additional elements that further describe the trait:

  • stat: If a trait describes a group of of measurements (as opposed to a single measurement), a stat element may be included. It must have the following three attributes:

    • sample_size, a positive integer.

    • name, the name of the statistic measured. Possible values are "SD", "SE", "MSE" "95%Cl", "LSD", and "MSD".

    • value, a double giving the value of the named statistic.

  • notes: This is an element having no attributes but containing free-form textual content.

• Using a single entity for the whole data set.

  If all of the traits in the data set should share the same entity, it is possible to specify this by placing an entity element as the first child of the root trait-data-set element. The element has the same form as an entity element contained inside a trait element except that this global entity is allowed to be anonymous, that is, to have no name or notes attribute. If a global entity element is used, it must be the only entity element in the document, and no trait-group elements may be used in the document (see below).

• Trait groups.

  If a group of traits share a number of characteristics, it is possible to nest them within a trait-group element. This is mainly useful in the following two cases:

  • Some (but not all) of the traits in the file should be associated with the same entity.

  • Some (but not all) of the traits in the file share the same metadata (site, citation, treatment, variable, date, species, etcetera).

  Multiple level of nesting may be used: trait-group elements may themselves contain trait-group elements.

• Entities for trait groups.

  If a trait-group element has no trait-group child element, then it may contain, as its first child element, an entity element. This usage is similar to the data-set entity usage describe above except that the entity will only be used for the traits in the trait group. If a trait group does use an entity element, then none of the traits in the trait group can have their own entity element.

• Specifying metadata for sets of traits.

  If many traits have a common citation, site, species, etcetera, it is possible to avoid repeating this information for each individual trait by using a defaults element. A defaults element may appear as the child of the trait-data-set element (in which case the defaults apply to all of the traits in the document) or as the child of a trait-group element (in which case it applies only to the traits within that group).

  defaults elements have many of the same attributes and child elements as trait elements:

  • Allowed attributes are access_level, utc_datetime, and local_datetime. local_datetime is allowed only if a site having a time zone is specified in the defaults element or in a defaults element at a higher level and if the specified site is not overridden at a lower level (see below).

  • Allowed child elements are site, species, citation, treatment, variable, and method.

  • As for the other attributes and elements used with trait elements, since the mean attribute and the stat, notes, and covariates elements are inherently trait-specific, they cannot be used with the defaults element. (entity elements applying to groups of traits are direct children of the trait-data-set element or a trait-group element rather than being nested within a defaults element.)

  A default specified by a defaults element will apply to all traits occuring within the parent of the defaults element unless overridden. A default may be overridden either by another defaults element appearing at a lower level or by attributes and child elements of an individual trait.

6.2.2 Schema for JSON Data Files

[To-do]

6.2.3 Schema for CSV Data Files

The format to use for CSV trait data files is largely the same as that required for the Bulk Upload wizard explained in the previous section. (See the templates traits.csv and traits_by_doi.csv.) Some significant differences from the bulk-upload case are:

1. The date of a trait measurement must be given in a column with one of the following headings.

   1. If the heading "utc_datetime" is used, the supplied values must conform to one of the following formats: 1918-11-11T10:00:00Z or 1918-11-11Z. In particular, the time must be given in UTC time (hence the "Z"), and if the time is specified (first format), the letter "T" must separate the date and the time portions. If the time is specified, seconds must be included; optionally, fractional seconds may be included as well. The resulting dateloc value will always be 5 (exact date); the timeloc value will be 1 (time to the second) if a time is given and 9 (no data) otherwise.

   2. If the heading "local_datetime" is used, the supplied values must conform to one of the following formats: 1918-11-11T11:00:00 or 1918-11-11. In particular, if the time is specified (first format), the letter "T" must separate the date and the time portions. If the time is specified, seconds must be included; optionally, fractional seconds may be included as well. The resulting dateloc value will always be 5 (exact date); the timeloc value will be 1 (time to the second) if a time is given and 9 (no data) otherwise.

      When "local_datetime" is used to specify the date-time value of a trait measurement, the date and time are assumed to be local (site) time if a site is given and if that site has a time zone value stored. Otherwise, the value given is assumed to be UTC time. (The date-time value is always stored in the database as UTC time. This paragraph has to do with how the supplied date-time value is interpreted when read.)

   Note that only one or the other of these columns may occur in the CSV file. Otherwise an error results.

2. Meta-data can not be specified interactively. Thus any associated citation, site, species, cultivar, or treatment must be specified in each row of the CSV file. (This may later change so that repeated metadata specification may be avoided.)

3. Unlike the bulk upload case, matching of metadata entries is case sensitive. Thus, if the CSV file specifies the species as "Sorghum Bicolor" but the database entry for the species specifies the scientificname as "Sorghum bicolor", the upload will not be successful.

4. Unlike the bulk upload case, it is not necessary to specify an associated citation, site, species, or treatment. The sample file SIMPLE_CSV_TEST_DATA demonstrates the case where a trait value having no associated metadata is inserted.

5. It is required, however, to specify an access level for each trait; therefore, the CSV file must have a column named access_level.

6. When specifying the citation in a CSV file for use with the Bulk Upload wizard, it is necessary to have either a citation_doi column, or have all three of the columns citation_author, citation_year, and citation_title. When using the API, however, any combination of these may be used so long as the values specified in each row determine a unique citation. For example, if there is only one citation with author "Doe", and if that is the value that occurs in the citation_author column of every row of the table, then it is unnecessary to have a citation_year or citation_title column.

7. Just as for bulk uploads, the trait_covariate_associations table is consulted to determine which column names correspond to trait variables and which ones correspond to covariate variables, and further, which covariates correspond to which traits. But failing to specify a required covariate for one or more traits will not result in an error. (Thus, in essence, required covariates are treated just like optional covariates; they will be associated if present, but no complaint will be made if they are not.)

8. No rounding is done of floating point values except to the extent required to fit within PostgreSQL's 8-byte float type. (Note that all floating point values may be specifed with an exponent; for example 3.20E-2 in place of 0.0320.)

6.3 Error Feedback

As mentioned above, all trait insertion API calls generate an HTTP response. The response will use the same format as the format of the file submitted except in the case of CSV files, where the response is given in JSON format.

In the case of unsuccessful API calls, the response will contain information about the types of errors that caused the call to be unsuccessful. These errors can be classified as follows:

6.3.1 Authorization Errors

If an invalid API key is given, or if the given key is for a user who isn't authorized to perform the given action, an authorization error is returned. (To do: Distinguish between authentication and authorization.)

6.3.2 Lookup Errors

If a citation, site, species, or treatment is specified that doesn't match exactly one item in the database, a lookup error occurs. This causes the whole data set to be returned in tree form as annotated_post_data. The annotations will be the error items next to the data item that caused the error.

6.3.3 Schema Validation Errors

As mentioned above, data files in CSV and JSON format are converted to XML format and then validated against an XML schema. For CSV files, since the structure of the XML document generated by the converter is generally correct, this error usually arises only when a data value of the wrong type is given (for example, an alphabetical string where a number is expected). But there are other situations that can trigger a validation error: for example, if a sample size column (n) is given without including a standard error (SE) column, or vice versa.¹²

6.3.4 Model Validation Errors

These errors occur when attempting to save a Trait object to the database and may occur if a variable value is found to be out of range or if a required attribute (e.g. access_level) is missing. As in the lookup error case, this causes the whole data set to be returned in tree form as annotated_post_data.

6.4 Running the Examples

There are five sample data files in the directory app/lib/api/test.

SIMPLE_XML_TEST_DATA: A minimal XML data file consisting of a single trait. It provides only the (currently) required values: the trait variable name, the trait value, and an access level to specify who may view this data item.

SIMPLE_CSV_TEST_DATA: A minimal CSV data file consisting of a single trait. It provides only a single variable name and value and the access level.

TEST_XML_DATA: A full-fledged XML data file making use of all of the features available for XML trait data insertion: specification of defaults for groups of traits, meta-data lookup, and complete latitude for associating specific groups of covariates and specific sample statistics with specific traits.

TEST_JSON_DATA: This JSON data file is an exact analogue to TEST_XML_DATA; it should result in exactly the same trait data being inserted.

TEST_CSV_DATA: This CSV data file has five column headings corresponding trait variables and two columns headings corresponding to covariate variable. There is a single data row, so when this file is ingested, a single new entity will be created having 5 associated traits and each trait will have 2 associated covariates. Complete metadata is given for the entity (or equivalently, for the traits it comprises).

6.4.1 Using curl to Upload the Data Samples

You can upload the data in these files using curl. To try this out, start your Rails server locally with rails s and then, from the /api/lib/api/test directory, run the command

curl -X POST --data-binary @TEST_XML_DATA localhost:3000/api/v1/traits.xml?key=<your API key>

(Substitute any of the other sample file names for TEST_XML_DATA as desired, changing the .xml extension to .json or .csv where appropriate. If a CSV file is being uploaded, add the option -H "Content-Type: text/csv" to the curl command.)

These API calls all generate a response (in XML format for the XML endpoint and in JSON format for the JSON endpoints). If the call is successful, the response will contain a list of the ids of the new traits that were inserted. Note that new entities and possibly new covariates will also be inserted, but the information about these is not (currently) contained in the response.

6.5 Known Bugs

1. It's too easy to make a mistake without realizing it.

   Examples:

   1. If you misspell a trait variable name in the heading, that column will simple be ignored; no error will occur if there exists at least one valid trait variable in the heading.

   2. If you include the same heading twice, the value is one column will overwrite those in the other.

2. Some error messages are obscure and seemingly unrelated to the error that triggered them.