Querying infinite parent depth for deep release entities.
This commit is contained in:
80
src/deep.js
80
src/deep.js
@@ -5,49 +5,11 @@ const merge = require('object-merge-advanced');
|
||||
|
||||
const argv = require('./argv');
|
||||
const include = require('./utils/argv-include')(argv);
|
||||
const { fetchReleaseEntities, urlToSiteSlug } = require('./entities');
|
||||
const logger = require('./logger')(__filename);
|
||||
const knex = require('./knex');
|
||||
const qu = require('./utils/qu');
|
||||
const scrapers = require('./scrapers/scrapers');
|
||||
|
||||
function urlToSiteSlug(url) {
|
||||
try {
|
||||
const slug = new URL(url)
|
||||
.hostname
|
||||
.match(/([\w-]+)\.\w+$/)?.[1]
|
||||
.replace(/[-_]+/g, '');
|
||||
|
||||
return slug;
|
||||
} catch (error) {
|
||||
logger.warn(`Failed to derive entity slug from '${url}': ${error.message}`);
|
||||
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
async function findEntities(baseReleases) {
|
||||
const baseReleasesWithoutEntity = baseReleases.filter(release => release.url && !release.site && !release.entity);
|
||||
|
||||
const entitySlugs = Array.from(new Set(
|
||||
baseReleasesWithoutEntity
|
||||
.map(baseRelease => urlToSiteSlug(baseRelease.url))
|
||||
.filter(Boolean),
|
||||
));
|
||||
|
||||
const entities = await knex('entities')
|
||||
.select(knex.raw('entities.*, row_to_json(parents) as parent, json_agg(children) as children'))
|
||||
.leftJoin('entities as parents', 'parents.id', 'entities.parent_id')
|
||||
.leftJoin('entities as children', 'children.parent_id', 'entities.id')
|
||||
.whereIn('entities.slug', entitySlugs)
|
||||
.groupBy('entities.id', 'parents.id')
|
||||
.orderBy('entities.type', 'asc');
|
||||
|
||||
// channel entity will overwrite network entity
|
||||
const entitiesBySlug = entities.reduce((accEntities, entity) => ({ ...accEntities, [entity.slug]: accEntities[entity.slug] || entity }), {});
|
||||
|
||||
return entitiesBySlug;
|
||||
}
|
||||
|
||||
function toBaseReleases(baseReleasesOrUrls, entity = null) {
|
||||
if (!baseReleasesOrUrls) {
|
||||
return [];
|
||||
@@ -106,8 +68,32 @@ async function fetchScene(scraper, url, entity, baseRelease, options) {
|
||||
return null;
|
||||
}
|
||||
|
||||
async function scrapeRelease(baseRelease, entities, type = 'scene') {
|
||||
const entity = baseRelease.entity || entities[urlToSiteSlug(baseRelease.url)];
|
||||
function findScraper(entity) {
|
||||
if (scrapers.releases[entity.slug]) {
|
||||
return scrapers.releases[entity.slug];
|
||||
}
|
||||
|
||||
if (entity.parent) {
|
||||
return findScraper(entity.parent);
|
||||
}
|
||||
|
||||
return null;
|
||||
}
|
||||
|
||||
function findLayoutScraper(entity, scraper) {
|
||||
if (scraper?.[entity.parameters?.layout]) {
|
||||
return scraper[entity.parameters.layout];
|
||||
}
|
||||
|
||||
if (entity.parent) {
|
||||
return findLayoutScraper(entity.parent, scraper);
|
||||
}
|
||||
|
||||
return scraper;
|
||||
}
|
||||
|
||||
async function scrapeRelease(baseRelease, entitiesBySlug, type = 'scene') {
|
||||
const entity = baseRelease.entity || entitiesBySlug[urlToSiteSlug(baseRelease.url)];
|
||||
|
||||
if (!entity) {
|
||||
logger.warn(`No entity available for ${baseRelease.url}`);
|
||||
@@ -121,8 +107,8 @@ async function scrapeRelease(baseRelease, entities, type = 'scene') {
|
||||
};
|
||||
}
|
||||
|
||||
const scraper = scrapers.releases[entity.slug] || scrapers.releases[entity.parent?.slug] || scrapers.releases[entity.parent?.parent?.slug];
|
||||
const layoutScraper = scraper?.[entity.parameters?.layout] || scraper?.[entity.parent?.parameters?.layout] || scraper?.[entity.parent?.parent?.parameters?.layout] || scraper;
|
||||
const scraper = findScraper(entity);
|
||||
const layoutScraper = findLayoutScraper(entity, scraper);
|
||||
|
||||
if (!layoutScraper) {
|
||||
logger.warn(`Could not find scraper for ${baseRelease.url}`);
|
||||
@@ -184,19 +170,19 @@ async function scrapeRelease(baseRelease, entities, type = 'scene') {
|
||||
}
|
||||
}
|
||||
|
||||
async function scrapeReleases(baseReleases, entities, type) {
|
||||
async function scrapeReleases(baseReleases, entitiesBySlug, type) {
|
||||
return Promise.map(
|
||||
baseReleases,
|
||||
async baseRelease => scrapeRelease(baseRelease, entities, type),
|
||||
async baseRelease => scrapeRelease(baseRelease, entitiesBySlug, type),
|
||||
{ concurrency: 10 },
|
||||
);
|
||||
}
|
||||
|
||||
async function fetchReleases(baseReleasesOrUrls, type = 'scene') {
|
||||
const baseReleases = toBaseReleases(baseReleasesOrUrls);
|
||||
const entities = await findEntities(baseReleases);
|
||||
const entitiesBySlug = await fetchReleaseEntities(baseReleases);
|
||||
|
||||
const deepReleases = await scrapeReleases(baseReleases, entities, type);
|
||||
const deepReleases = await scrapeReleases(baseReleases, entitiesBySlug, type);
|
||||
|
||||
return deepReleases.filter(Boolean);
|
||||
}
|
||||
|
||||
@@ -66,6 +66,21 @@ async function curateEntities(entities, includeParameters) {
|
||||
return Promise.all(entities.map(async entity => curateEntity(entity, includeParameters)));
|
||||
}
|
||||
|
||||
function urlToSiteSlug(url) {
|
||||
try {
|
||||
const slug = new URL(url)
|
||||
.hostname
|
||||
.match(/([\w-]+)\.\w+$/)?.[1]
|
||||
.replace(/[-_]+/g, '');
|
||||
|
||||
return slug;
|
||||
} catch (error) {
|
||||
logger.warn(`Failed to derive entity slug from '${url}': ${error.message}`);
|
||||
|
||||
return null;
|
||||
}
|
||||
}
|
||||
|
||||
async function fetchIncludedEntities() {
|
||||
const include = {
|
||||
includeAll: !argv.networks && !argv.channels && !config.include?.networks && !config.include?.channels,
|
||||
@@ -139,6 +154,46 @@ async function fetchIncludedEntities() {
|
||||
return curatedNetworks;
|
||||
}
|
||||
|
||||
async function fetchReleaseEntities(baseReleases) {
|
||||
const baseReleasesWithoutEntity = baseReleases.filter(release => release.url && !release.site && !release.entity);
|
||||
|
||||
const entitySlugs = Array.from(new Set(
|
||||
baseReleasesWithoutEntity
|
||||
.map(baseRelease => urlToSiteSlug(baseRelease.url))
|
||||
.filter(Boolean),
|
||||
));
|
||||
|
||||
const entities = await knex.raw(`
|
||||
WITH RECURSIVE tree as (
|
||||
SELECT to_jsonb(entities) as entity,
|
||||
parent_id,
|
||||
array['parent'] as parent_path,
|
||||
0 as depth
|
||||
FROM entities
|
||||
WHERE slug = ANY(:entitySlugs)
|
||||
|
||||
UNION ALL
|
||||
|
||||
SELECT jsonb_set(tree.entity, tree.parent_path, to_jsonb(entities)),
|
||||
entities.parent_id,
|
||||
tree.parent_path || array['parent'],
|
||||
depth + 1
|
||||
FROM tree
|
||||
JOIN entities ON tree.parent_id = entities.id
|
||||
)
|
||||
SELECT entity FROM tree WHERE parent_id is null
|
||||
ORDER BY entity->'type' ASC;
|
||||
`, { entitySlugs });
|
||||
|
||||
// channel entity will overwrite network entity
|
||||
const entitiesBySlug = entities.rows.reduce((accEntities, { entity }) => ({
|
||||
...accEntities,
|
||||
[entity.slug]: accEntities[entity.slug] || curateEntity(entity, true),
|
||||
}), {});
|
||||
|
||||
return entitiesBySlug;
|
||||
}
|
||||
|
||||
async function fetchEntity(entityId, type) {
|
||||
const entity = await knex('entities')
|
||||
.select(knex.raw(`
|
||||
@@ -290,8 +345,10 @@ module.exports = {
|
||||
curateEntity,
|
||||
curateEntities,
|
||||
fetchIncludedEntities,
|
||||
fetchReleaseEntities,
|
||||
fetchEntity,
|
||||
fetchEntities,
|
||||
searchEntities,
|
||||
flushEntities,
|
||||
urlToSiteSlug,
|
||||
};
|
||||
|
||||
@@ -164,7 +164,6 @@ function attachReleaseIds(releases, storedReleases) {
|
||||
|
||||
function filterInternalDuplicateReleases(releases) {
|
||||
const releasesByEntityIdAndEntryId = releases.reduce((acc, release) => {
|
||||
console.log(release);
|
||||
if (!release.entity) {
|
||||
return acc;
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user