Skip to content

Latest commit

 

History

History
74 lines (43 loc) · 3.08 KB

PIPELINES.md

File metadata and controls

74 lines (43 loc) · 3.08 KB
Raw

Pipelines

After an Item is emitted from a spider, it goes through several pipelines before it is exported. If a spider needs to disable these pipelines, they can by overriding ITEM_PIPELINES in custom_settings. See fedex.py for an example.

DropAttributesPipeline

When a spider has drop_attributes, we remove those attributes from Features, this happens early, so it may be re added by other pipelines later on.

ApplySpiderLevelAttributesPipeline

When a spider has item_attributes, they are applied to all the Items produced by the Spider. We don't override existing attributes, so the spider is free to override its own item_attributes. This typically includes brand details, however it can be any Item field as well as extras where we again only override fields that haven't been set in spider.

CountryCodeCleanUpPipeline

This normalises the country attribute to a 2-letter country code. It first checks if a country name has been provided, and if it can turn it onto a country code. When this fails or no country has been provided, we look at the Item to work out the country code. This can be disabled with a spider level attribute skip_auto_cc = True.

  1. Spider name: when the spider name ends with a single country code we apply that.

    This can be disabled by setting skip_auto_cc_spider_name = True in the Spider.

  2. Domain name: when website has a country level tld we use it to set the country code.

    This can be disabled with skip_auto_cc_domain = True.

  3. Geocoding: when the above fail, we try to geocode the Item using its latitude and longitude.

    This can be disabled with skip_auto_cc_geocoder = True.

StateCodeCleanUpPipeline

Items in Canada or the United States get state cleaned up to a 2-letter state code.

PhoneCleanUpPipeline

Phone numbers and phone number-like extra attributes (fax, *:phone and *:fax) are normalized to an international format. We currently output the original data from spider if we fail to parse it.

ExtractGBPostcodePipeline

We attempt to extract the postcode from Items in Ireland or Great Britain that don't have postcode but do have addr_full.

AssertURLSchemePipeline

We apply the https scheme to Items with image that are missing a URL scheme.

DropLogoPipeline

Items with image that mentions logo or favicon get dropped.

ClosePipeline

Count Items that are flagged as closed and log other Items that may be closed. We do this for gathering statistics and debugging.

ApplyNSICategoriesPipeline

Attempt to match Items against Name Suggestion Index (NSI). When successful, we apply the attributes from NSI to the Item, not overriding the spider-provided attributes.

See NSI Matching for more details.

This can be disabled by setting nsi_id='N/A' in the Spider.

CheckItemPropertiesPipeline

Count missing and invalid properties provided by the spider. This pipeline will drop coordinate at and near Null Island, however it won't drop other invalid properties.