refacto get datasets beta.gouv

.env.sample

@@ -6,6 +6,7 @@ ADMIN_TOKEN=
 API_DEPOT_URL=https://plateforme-bal.adresse.data.gouv.fr/api-depot
 API_DEPOT_CLIENT_SECRET=
 API_DEPOT_CLIENT_ID=
+URL_API_DATA_GOUV=https://www.data.gouv.fr/api/1
 
 S3_ENDPOINT=https://s3.gra.io.cloud.ovh.net/
 S3_CONTAINER_ID=

lib/sources/datagouv.js

@@ -1,57 +1,42 @@
 const got = require('got')
 const {chain} = require('lodash')
 
-function isCertified(organization) {
-  const {badges} = organization
-
-  return badges.some(b => b.kind === 'certified')
-    && badges.some(b => b.kind === 'public-service')
-}
-
-function isBAL(resource) {
-  return resource.format === 'csv' || resource.url.endsWith('csv')
-}
+const URL_API_DATA_GOUV = process.env.URL_API_DATA_GOUV || 'https://www.data.gouv.fr/api/1'
+const PAGE_SIZE = 100
+const TAG = 'base-adresse-locale'
+const FORMAT = 'csv'
 
+// CREATE INTERNE CACHE
 const organizationsCache = {}
+const datasetsCache = {}
 
 async function getOrganization(organizationId) {
   if (!(organizationId in organizationsCache)) {
-    const response = await got(`https://www.data.gouv.fr/api/1/organizations/${organizationId}/`, {responseType: 'json'})
+    const response = await got(`${URL_API_DATA_GOUV}/organizations/${organizationId}/`, {responseType: 'json'})
     organizationsCache[organizationId] = response.body
   }
 
   return organizationsCache[organizationId]
 }
 
-const datasetsCache = {}
-
 async function getDataset(datasetId) {
   if (!(datasetId in datasetsCache)) {
-    const response = await got(`https://www.data.gouv.fr/api/1/datasets/${datasetId}/`, {responseType: 'json'})
+    const response = await got(`${URL_API_DATA_GOUV}/datasets/${datasetId}/`, {responseType: 'json'})
     datasetsCache[datasetId] = response.body
   }
 
   return datasetsCache[datasetId]
 }
 
-async function getEligibleBALDatasets() {
-  const response = await got('https://www.data.gouv.fr/api/1/datasets/?tag=base-adresse-locale&format=csv&page_size=1000', {responseType: 'json'})
-
-  // Register in datasets cache
-  response.body.data.forEach(dataset => {
-    datasetsCache[dataset.id] = dataset
-  })
+function isCertified(organization) {
+  const {badges} = organization
 
-  const datasets = await Promise.all(
-    response.body.data
-      .filter(d => d.resources.some(r => isBAL(r)) && d.organization && !d.archived)
-      .map(async d => {
-        const organization = await getOrganization(d.organization.id)
-        return {...d, organization}
-      })
-  )
+  return badges.some(b => b.kind === 'certified')
+    && badges.some(b => b.kind === 'public-service')
+}
 
-  return datasets.filter(d => isCertified(d.organization))
+function isBAL(resource) {
+  return resource.format === 'csv' || resource.url.endsWith('csv')
 }
 
 function getBALUrl(dataset) {
@@ -64,4 +49,39 @@ function getBALUrl(dataset) {
   return mostRecentResource.url
 }
 
+function computeBetaGouvDatasetsUrl(page) {
+  return URL_API_DATA_GOUV
+    + '/datasets/'
+    + `/?tag=${TAG}`
+    + `&format=${FORMAT}`
+    + `&page_size=${PAGE_SIZE}`
+    + `&page=${page}`
+}
+
+async function fetchDatasets(page = 1) {
+  const url = computeBetaGouvDatasetsUrl(page)
+  const response = await got(url, {responseType: 'json'})
+
+  // FILTER DATASETS
+  const datasets = response.body.data
+    .filter(d => d.resources.some(r => isBAL(r)) && d.organization && !d.archived && isCertified(d.organization))
+
+  if (response.body.total > page * PAGE_SIZE) {
+    return [...datasets, ...await fetchDatasets(page + 1)]
+  }
+
+  return datasets
+}
+
+async function getEligibleBALDatasets() {
+  // GET DATASET
+  const datasets = await fetchDatasets()
+  // BUILD CACHES
+  datasets.forEach(dataset => {
+    datasetsCache[dataset.id] = dataset
+  })
+
+  return datasets
+}
+
 module.exports = {getEligibleBALDatasets, getOrganization, getDataset, getBALUrl}

lib/sources/index.js

@@ -18,15 +18,16 @@ async function augmentCustomEntry(entry) {
     organization = await getOrganization(entry.organization || dataset.organization.id)
   }
 
+  // ADD URL TO ENTRY
   if (!entry.url && !entry.converter && dataset) {
-    return {...entry, dataset, organization, url: getBALUrl(dataset)}
+    entry.url = getBALUrl(dataset)
   }
 
   return {
-    ...entry,
     type: 'github',
     dataset,
-    organization
+    organization,
+    ...entry
   }
 }
 
@@ -73,13 +74,16 @@ function computeMetaFromSource(source) {
 async function computeList() {
   const customSources = readYamlFile(sourcesFilePath)
 
+  // CREATE BLACKLIST ID DATASET
   const blacklistedIds = customSources.blackList.map(e => e.dataset)
+  // CREATE WHITELIST ID DATASET
   const whitelistedIds = customSources.whiteList.filter(e => e.dataset).map(e => e.dataset)
-
+  // GET DATASETS FROM DATAGOUV
   const eligibleBALDatasets = await getEligibleBALDatasets()
+  // FILTER DATASETS WITHOUT BLACKLISTE AND WHITELISTE
   const selectedDatasets = eligibleBALDatasets
     .filter(d => !blacklistedIds.includes(d.id) && !whitelistedIds.includes(d.id))
-
+  // MIX WHITELIST AND DATAGOUV DATASETS
   const sources = [
     ...(await Promise.all(customSources.whiteList.map(source => augmentCustomEntry(source)))),
     ...(selectedDatasets.map(dataset => prepareEligibleEntry(dataset)))