We’re interested in the "purge config to always leave the most recent copy of any purge form type" as well. Unfortunately, this is the only mention of such functionality we found while combing through the CHT forum threads.
Similar to the original post, we want to keep the last report to ensure that a subsequent form can load previous values via the passed-in context (provided through the contact summary). For something like legitimate follow ups.
At the moment, we only have time-based purging, which, to quote our site manager, is a dealbreaker if it compromises the continuity of care data. For that reason, we can’t risk rolling out the purge script to CHWs just yet.
As far as I know, the purge script loops through documents one by one. Since it’s a self-contained function, it’s unable to “check” against other documents to determine if the current item is the newest. @diana would that be trivial to add?
Our purge script content:
// As of 3.14.0, contacts that have more than 20,000 associated reports + messages will be skipped,
// and none of their associated reports and messages will be purged.
// NOTE: Purging does not touch documents in the medic database, everything is done in separate purge databases(medic-purged-roles-<roles-hash>).
// A purgelog document is saved in the medic-sentinel database after every purge. The purgelog has a meaningful id: purgelog:<timestamp>, where timestamp represents the moment when purging was completed.
// Errors can be found in the purgelog as purgelog:error:<timestamp>
// https://docs.communityhealthtoolkit.org/apps/guides/performance/purging/#considerations
// ############################################### PURPOSE ###############################################
// We've been experiencing replication issues on the topmost users (DHO, Team Lead).
// The purge script will reduce the number of records synced to the client devices of these login users.
module.exports = {
// We need to find a sweet spot, as firing too frequently could overburden our already modestly specced server
// Purging may take more than 31 hours - https://forum.communityhealthtoolkit.org/t/purging-on-3-14-2-after-8-hours-waiting-how-can-i-know-if-purging-is-still-in-progress/1837/2
// https://forum.communityhealthtoolkit.org/t/purging-on-3-14-2-after-8-hours-waiting-how-can-i-know-if-purging-is-still-in-progress/1837/2
run_every_days: 7, // The interval (in days) at which purges will be downloaded client-side. Default 7.
// 'text_expression': 'at 11:00 pm on Fri', // Any valid text expression to describe the interval of running purge server-side. For more information, see https://bunkat.github.io/later/parsers.html#text
cron: '0 23 * * 5', // Same as above, just a different syntax. Either can be used, but one is required.
// For more info on how to set the intervals see: https://docs.communityhealthtoolkit.org/building/guides/performance/purging/#schedule-configuration
// The cron strings can be tested here: https://crontab.guru/
/* eslint-disable-next-line no-unused-vars */ // Provides more info about the available tie-ins
fn: function (userCtx = {}, contact = {}, reports = [], messages, chtScriptApi, permissions) {
// NOTE: the purge function CAN purge contacts, but it does not purge linked children.
// This means that there can be dangling records replicated and could still impact performance.
// We could consider having a static list of "top-ish level place IDs", and then remove any record with such a parent_id.
// That could be very cumbersome to maintain, and the contacts probably has a way smaller impact on performance than app forms.
const NOW = Date.now();
// The purge function is self-contained. See linked CHT docs above
// This method could add an overhead of between 0 - 3 days depending on the month, per month.
// It will ensure all records for the given months are retrieved.
const monthsAgo = months => NOW - 1000 * 60 * 60 * 24 * 31 * months;
// NOTE: Make sure this is kept up to date as the roles & responsibilities of the app evolve over time!
const householdCOPC = 'copc-hhscreening';
const householdCSharp = 'csharp-householdconsentandquestionnaire';
const TEST_FORM = 'YYYZ'; // This form is only used when testing the purge script for new roles
const individualCOPC = 'copc-individualhealthcaretasks';
const individualCSharp = 'csharp-individualhhconsentedquestionnaire';
const individualDeath = 'death_report';
// The "appliesIf" property could perform some sort of calculation to check if the purge applies. Default true.
// Three "preserve" "types" can be defined: -1 = all, 0 = none, any other number of months
// The same config can be applied to report_ and contact_types.
const PRESERVE_ALL = -1;
const PRESERVE_NONE = 0;
const CONF = Object.freeze({
// TODO: dho
// 'vap': {
// // We don't care about any reports except for the death report
// 'reportTypes': {
// [householdCOPC] : PRESERVE_NONE,
// [householdCSharp]: PRESERVE_NONE,
// [individualCOPC]: PRESERVE_NONE,
// [individualCSharp]: PRESERVE_NONE,
// [individualDeath]: 3
// },
// // We also don't care about any individuals that aren't marked for death
// 'contactTypes': {
// // We can't delete team_areas, indawos, dwellings, or households as we don't know which contains flagged individuals
// 'hhm': {
// 'appliesIf': (doc) => !doc['death_flag'] || doc['death_flag'] === '' || doc['death_flag'] === 'no',
// 'preserve': 3
// },
// // NOTE: We can, however, remove all other hierarchy persons EXCEPT our own!
// // If you don't take care the app will become unusable. It's a terrible experience when you delete yourself
// 'dho': PRESERVE_NONE,
// 'team_lead': PRESERVE_NONE,
// 'chw': PRESERVE_NONE
// }
// },
'team_lead': {
'reportTypes': {
[householdCOPC] : 1,
[householdCSharp]: 1,
[individualCOPC]: 1,
[individualCSharp]: 1,
[individualDeath]: 1,
[TEST_FORM]: PRESERVE_NONE,
},
}
// TODO: chw
});
// TODO: we have not considered login users with multiple roles!
const role = userCtx && userCtx.roles && userCtx.roles.length >= 1 ? userCtx.roles[0] : false;
if (role && role in CONF) {
const reportTypes = CONF[role]['reportTypes'];
const contactTypes = CONF[role]['contactTypes'];
const shouldPurge = (doc, conf) => {
if(typeof conf === 'object'){
return (conf['appliesIf'] ? conf['appliesIf'](doc) : true) && doc.reported_date <= monthsAgo(conf['preserve']);
}
else if(typeof conf === 'number'){
return conf !== PRESERVE_NONE ? conf !== PRESERVE_ALL? doc.reported_date <= monthsAgo(conf) : false : true;
}
return false;
};
const purgeContact = (contactTypes && contact.contact_type in contactTypes && shouldPurge(contact, contactTypes[contact.contact_type]) ? [contact._id] : []).filter((value) => value);
const reportsToPurge = reports.filter((doc) => doc.form in reportTypes && shouldPurge(doc, reportTypes[doc.form])).map(r => r._id).filter((v) => v);
// Purging of messages is not needed as we don't use the feature
return [
...purgeContact,
...reportsToPurge
];
}
return []; // Do not purge anything
}
};
// Because JSON is fantastic with comments, we have to add the `forms.json` content here
// This can be used to allow forms posted via the CHT "/api/v2/records" API
// {
// "YYYZ": {
// "meta": {
// "code": "YYYZ"
// },
// "fields": {
// "patient_id": {
// "labels": {
// "short": {
// "translation_key": "form.flag.patient_id.short"
// },
// "tiny": "pid"
// },
// "position": 0,
// "type": "string",
// "length": [
// 5,
// 13
// ],
// "required": true
// }
// },
// "public_form": true
// }
// }
// The post request body that adheres the above config & api:
// {
// "patient_id": "<your_patient_id_here>",
// "form": "YYYZ",
// "nurse": "Sam",
// "week": 23,
// "year": 2015,
// "visit": "ANC",
// "fields": {
// "patient_id": "<your_patient_id_here>"
// },
// "_meta": {
// "form": "YYYZ",
// "reported_date": 1725001661000
// }
// }
I do wonder if it is possible to setup your purge config to always leave the most recent copy of any purged form type. 
If not, that may be a possible feature to pursue adding…