I don't get why all of the big RDBMSes (PostgreSQL, MariaDB/MySQL, SQL Server, Oracle, ...) don't seem to have built in support for soft deletes up front and center?
CREATE TABLE ... WITH SOFT DELETES
Where the regular DELETE wouldn't get rid of the data for real but rather you could query the deleted records as well, probably have timestamps for everything as a built in low level feature, vs having to handle this with a bunch of ORMs and having to remember to put AND deleted_at IS NULL in all of your custom views.
If we like to talk about in-database processing so much, why don't we just put the actual common features in the DB, so that toggling them on or off doesn't take a bunch of code changes in app, or that you'd even be able to add soft deletes to any legacy app that knows nothing of the concept, on a per table basis or whatever.
Maybe my intuition is wrong, but to me this sounds like a violation of the principle of least power.
"Soft deletes" is just a name for a regular write operation, with specific semantics.
Adding a layer of magic to the DB for this doesn't seem right to me.
And applications could have many different requirements for soft deletes, like the article points out. For example, the simplest version would be just a boolean "deleted" column, but it could also be "deleted_at", "deleted_by", etc.
All of these cases require an bunch of code changes anyway, and the more complex ones could interfere with an implementation of this feature at the database level: such a transparent implementation couldn't access app-specific concerns such as user data, for example.
Adding soft deletes to a legacy app that knows nothing about it would only work for a boolean flag and a maybe date-time value, unless the DBMS would also offer triggers for soft deletes etc?
Seems to me to that this capability would make a DBMS much more complicated.
These are not decisions that should be taken solely by whoever is programming the backend.
They need to be surfaced to the product owner to decide. There may very well be reasons pieces of data should not be stored. And all of this adds complexity, more things to go wrong.
If the product owner wants to start tracking every change and by who, that can completely change your database requirements.
So have that conversation properly. Then decide it's either not worth it and don't add any of these "extra" fields you "might" need, or decide it is and fully spec it out and how much additional time and effort it will be to do it as a proper feature. But don't do it as some half-built just-in-case "favor" to a future programmer who may very well have to rip it out.
On a personal project, do whatever you want. But on something professional, this stuff needs to be specced out and accounted for. This isn't a programming decision, it's a product decision.
This entirely depends on the company culture. I worked in teams where every small decision is in the hand of the PO and I've worked in teams where a software engineer is a respected professional enabled to make their own technical decisions. I found the second option to create higher quality software faster.
Also not sure what you mean by additional effort? Created_at, updated_at or soft-deletes are part of most proper frameworks. In Spring all you need is an annotation, I've been using those in major projects and implementation cost is around a few seconds with so far zero seconds of maintenance effort in years of development. At least those fields are solved problems.
Some things are trivial and nearly free - created_at, updated_at. I don't think engineers need to bring trivialities like this to a "product owner". Own your craft.
When the product you're developing is governed by regulations and standards you need to comply, owning your craft is doing things by the book, not adding fields on your own because it might be useful later.
Although those can be more complicated, and it should be clear what they're for and why they exist. Will this result in an object having an updated_at timestamp elsewhere in a larger application? Is it clear which properties that refers to?
I think the tricky part lies on knowing which things can be done without consulting any product owner. I agree that created_at and updated_at don’t cause any harm. deleted_at on the other hand cannot be decided by engineers only (mainly because of GDPR reasons: if something is expected to be totally deleted, then that must be it). As usual, these kind of things are obvious to engineers with years of experience , not so much to newcomers.
> And all of this adds complexity, more things to go wrong
That's a little vague given this specific example, which appears to be about maintaining some form of informative logging; though I don't think it necessarily needs to be in the form of an DB table.
I kinda agree, but don’t underestimate the power of having things where people are looking.
Put your documentation in doc strings where the function is defined - don’t have a separate file in a separate folder for that. It might separate concerns, but no one is looking there.
Similarly if those fields aren’t nullable, someone trying to add new rows will have to fill in something for those metadata fields - and that something will now very likely be what’s needed, rather than not pushing anything to the audit table.
Obviously your app can outgrow these simple columns, but you’re getting value now.
An audit log table often takes a huge amount of space compared to simple fields on the records so there are tradeoffs. Which solution is best depends on how important change logs are.
Event sourcing also works great. You don't need an audit log per se if you already track a history of all commands that introduced changes to your system.
Yep. But Event Sourcing comes with its own set of other problems.
Also, I don't think this would apply to OP's post: with Event Sourcing you would not even have those DB tables.
*_at and *_by fields in SQL are just denormalization + pruning patterns consolidated, right?
Do the long walk:
Make the schema fully auditable (one record per edit) and the tables normalized (it will feel weird). Then suffer with it, discover that normalization leads to performance decrease.
Then discover that pruned auditing records is a good middle ground. Just the last edit and by whom is often enough (ominous foreshadowing).
Fail miserably by discovering that a single missing auditing record can cost a lot.
Blame database engines for making you choose. Adopt an experimental database with full auditing history. Maybe do incremental backups. Maybe both, since you have grown paranoid by now.
Discover that it is not enough again. Find that no silver bullet exists for auditing.
Now you can make a conscious choice about it. Then you won't need acronyms to remember stuff!
Fair enough, but now your application is relying on 100% uptime of AWS and S3 and no network failures in between. And what happens if your transaction goes through, but the request to AWS doesn’t? What happens if another operation mutates the target meanwhile before you can retry with current state? Your app is also slowing down since it needs to send the events to S3 and guarantee they got there. Now you are reinventing two-stage commits. Unless you aren’t actually making an audit log and don’t care if events are guaranteed to be logged?
So like OP said, no silver bullets exist for auditing.
It is interesting thinking about record changes as a spectrum towards application logs. At some point too much detail is expensive to store, and you must adopt an archival strategy.
If you see it from the pure SQL point of view, you are in the "blame database engines and adopt an experimental solution".
It is the point where you give up modeling the audit as part of the systems tables.
The drawbacks of this choice are often related to retrieval. It depends on the engine.
I once maintained a system that kept a fully working log replicated instance delayed by 24h, ready for retrieval queries, in addition to regular disk backups (slow costy retrieval).
I am more developer than DBA, so I can probably speak more about modeling solutions than infra-centric solutions.
The problem with this is the audit log is only at the CRUD level which is often too low. Ambiguities can arise. For example if the question is "who published the article" do you look for a create or do you look for an update with published=true? It's even worse when you consider the schema can change over time, so both can be correct but at different points in time. Event sourcing is the way if you want to capture business-level events.
Additionally, mutable fields will quite often benefit from having a separate edit table which records the old value, the new value, who changed it, and when. Your main table’s created and updated times can be a function of (or a complement to) the edit table.
It is tempting to supernormalize everything into the relations object(id, type) and edit(time, actor_id, object_id, key, value). This is getting dangerously and excitingly close to a graph database implemented in a relational database! Implement one at your peril — what you gain in schemaless freedom you also lose in terms of having the underlying database engine no longer enforcing consistency on your behalf.
> This is getting dangerously and excitingly close to a graph database implemented in a relational database!
This feels like a great unresolved tension in database / backend design - or maybe I'm just not sophisticated enough to notice the solutions?
Is the solution event sourcing and using the relational database as a "read model" only? Is that where the truly sophisticated application developers are at? Is it really overkill for everybody not working in finance? Or is there just not a framework that's made it super easy yet?
Users demand flexible schemas - should we tell them no?
> Additionally, mutable fields will quite often benefit from having a separate edit table which records the old value, the new value, who changed it, and when.
Aren't you describing a non-functional approach to event sourcing? I mean, if the whole point of your system is to track events that caused changes, why isn't your system built around handling events that cause changes?
> supernormalize everything into the relations object(id, type) and edit(time, actor_id, object_id, key, value)
I frankly hate this sort of thing whenever I see it. Software engineers have a tendency to optimize for the wrong things.
Generic relations reduce the number of tables in the database. But who cares about the number of tables in the database? Are we paying per table? Optimize for the data model actually being understandable and consistently enforced (+ bonus points for ease of querying).
One thing I do quite frequently which is related to this (and possibly is a pattern in rails) is to use times in place of Booleans.
So is_deleted would contain a timestamp to represent the deleted_at time for example. This means you can store more information for a small marginal cost. It helps that rails will automatically let you use it as a Boolean and will interpret a timestamp as true.
I consider booleans a code smell. It's not a bug, but it's a suggestion that I'm considering something wrong. I will probably want to replace it with something more meaningful in the future. It might be an enum, a subclass, a timestamp, refactoring, or millions of other things, but the Boolean was probably the wrong thing to do even if I don't know it yet.
The way I think about it: a boolean is usually an answer to a question about the state, not the state itself.
A light switch doesn't have an atomic state, it has a range of motion. The answer to the question "is the switch on?" is a boolean answer to a question whose input state is a range (e.g. is distance between contacts <= epsilon).
This seems at first like a controversial idea, but the more I think about it the more I like this thought technology. Merely the idea of asking myself if there's a better way to store a fact like that will potentially improve designs.
32-bit UNIX timestamps are often signed so you can actually go before that, but most UNIX timestamps are 64-bit now, which can represent quite a larger range. And SQL datetime types might have a totally different range.
Not that it really matters; deleted_at times for your database records will rarely predate the existence of said database.
« Anytime you store Boolean, a kitten dies »
Nobody has ever said that but nobody wants any kitten to die so nobody has ever challenged me anytime I use that statement.
A little while back, I had a conversation with a colleague about sorting entries by "updated at" in the user interface, and to my surprise this was not added by the backend team.
Many of these "we are going to need it"s come from experience. For example in the context of data structures (DS), I have made many "mistakes" that I do correctly a second time. These mistakes made writing algorithms for the DS harder, or made the DS have bad performance.
Sadly, it's hard to transfer this underlying breadth of knowledge and intuition for making good tradeoffs. As such, a one-off tip like this is limited in its usefulness.
Database schemas being perfect out-of-the gate was replaced by reliable migrations.
If it's not data that's essential to serving the current functionality, just add a column later. `updated_at` doesn't have to be accurate for your entire dataset; just set it to `NOW()` when you run the migration.
Sure, migrations are bearable (especially ones that only add columns).
But for the example of the "updated_at" column, or "soft delete" functionality, you only find out you need it because the operations team suddenly discovered they needed that functionality on existing production rows because something weird happened.
In C#-land, we just have it as a standard that ~every table inherits from `ITrackable`, and we wrote a little EF plugin to automatically update the appropriate columns.
“Reliable migrations” almost seems like an oxymoron. Migrations are complicated, difficult and error prone. I think there’s a good takeaway here around good initial schema design practices. The less you have to morph your schema overtime, the less of those risky migrations need to run.
Still depends on what the data represent: you could get yourself in a storm of phone calls from customers if after your latest release there's now a weird note saying their saved document was last updated today.
Somewhat related, but I suggest having both the record updated at, and some kind of "user editing updated at". As I've encountered issues where some data migration ends up touching records and bumping the updated at, which shocks users since they see the UI reshuffle and think they have been hacked when they see the records updated at a time they didn't update them.
I like the heuristics described here.
However if these things aren't making it into a product spec where appropriate, then I smell some dysfunction that goes beyond what's being stored by default.
Product need (expressed as spec, design, etc) should highlight the failure cases where we would expect fields like these to be surfaced.
I'd hope that any given buisness shouldn't need someone with production database access on hand to inform as to why/when/how 'thing' was deleted. Really we'd want the user (be it 'boss' or someone else) to be able to access that information in a controlled manner.
"What information do we need when something goes wrong?". Ask it. Drill it. Ask it again.
That said, if you can't get those things, this seems a fine way to be pragmatic.
Yes! Why something happened is incredibly important. Gitlab made this mistake hard. We have a medium sized instance with some complex CI pipelines and often they'll just get cancelled and it doesn't say why or even who by. And anyone can do it! The only option is to ask the entire company "did anyone cancel this?"
I agree with this as written, as think it's important to have some degree of forethought when building out the DB to plan for future growth and needs.
That said, the monkey paw of this would be someone reading it and deciding they should capture and save all possible user data, "just in case", which becomes a liability.
Just curious, how do people feel about this general style of soft deletes currently? Do people still use these in production or prefer to just delete fully or alternatively move deleted rows to a separate tables / schema?
I find the complexity to still feel awkward enough that makes me wonder if deleted_at is worth it. Maybe there are better patterns out there to make this cleaner like triggers to prevent deletion, something else?
As for the article, I couldn't agree more on having timestamps / user ids on all actions. I'd even suggest updated_by to add to the list.
Financial world: records have a "close" or "expire" date which is then purged after some period of time. A deletion doesn't just happen, the record is updated to be "closed" or "expired" and some time after that it's deleted.
Something like a loan could live in a production environment for well over a year after closing, while an internal note may last just a month.
I think soft deletes using timestamptz are a good thing.
Deleting rows directly could mean you're breaking references. For example, say you have a product that the seller wants to delete. Well, what happens if customers have purchased that product? You still want it in the database, and you still want to fulfill the orders placed.
Your backend can selectively query for products, filter out deleted_at for any customer facing queries, but show all products when looking at purchase history.
There are times when deleting rows makes sense, but that's usually because you have a write-heavy table that needs clearing. Yes, soft-deletes requires being careful with WHERE statements filtering out deleted rows, but that's a feature not a bug.
> what happens if customers have purchased that product? You still want it in the database, and you still want to fulfill the orders placed.
You might still want to show to those customers their purchase history including what they bought 25 years ago. For example, my ISP doesn't have anymore that 10 Mb/s fiber optic product I bought im 2000, because it was superseded by 100 Mb/s products and then by 1 Gb/s ones. It's also not my ISP anymore but I use it for the SIM in my phone. That also accumulated a number of product changes along the years.
And think about the inventory of eshops with a zillion products and the archive of the pady orders. Maybe they keep the last few years, maybe everything until the db gets too large.
There can be legal requirements to retain data for a specified time for law enforcement and audits, while at the same time other legal requirements have you requiring to delete data upon customer request.
Doing this with pure 'hard' deletes is not possible, unless you maintain 2 different tables, one of which would still have the soft delete explicit or implicit. You could argue the full db log would contain the data for the former requirement, but while academicly correct this does not fly in practice.
If you have a good audit log, it really doesn't matter. You can always restore it if need be.
If you have no audit log(or a bad one), like lots of apps, then you have to care a lot.
Personally, I just implement a good audit log and then I just delete with impunity. Worst case scenario, someone(maybe even me) made a mistake and I have to run undo_log_audit() with the id of the audit log entry I want to put back. Nearly zero hassle.
The upside, when something goes wrong, I can tell you who, what and when. I usually have to infer the why, or go ask a human, but it's not usually even difficult to do that.
Always soft-deletion first. Then it gets exported to a separate archive and only then, after some time and may be attempted to be fully deleted from the initial base.
(It Is Probable That While Not Immediately Required The Implementation of Storage of Data In Question May Be Simpler Now Rather Than Later)
I've gone ahead and included additional detail in the acronym in the event that the clarity is required later, as this would be difficult to retrofit into a shorter, more-established acronym.
It's a terrible post. What it suggests is to turn your head off and follow overly generalised principle. I guess when somebody invent yet another acronym it is a red flag.
Data has its own life cycles in every area it passes through. And it's part of requirements gathering to find those cycles: the dependent systems, the teams, and the questions you need to answer. Mindlessly adding fields won't save you in every situation.
Bonus point: when you start collecting questions while designing your service, you'll discover how mature your colleagues' thinking is.
Five years ago everybody would lough about "soft deletes" or "marked as deleted". Whoever thought this is a good idea from a data protection perspective?
You also lying in the face of your users with such a behavior. Shame.
Except almost every database (and most storage devices nowadays) works this way - mark an entry as deleted, then batch delete a lot of entries during garbage collection. It's fundamentally impossible to efficiently erase a record from the middle of a file, except maybe by using an encryption tree, which would still be fairly inefficient.
Actually erasing data is quite hard. Soft deletes doesn't add any new lies, they just move the lies to the upper layer.
Come on. With a manual "marked as deleted" it stays as this forever, it is not deleted and never will, and the "deleted" data lands also in database backups, is still query-able and so on. I do not care if the deleted data stays for a while on disk or in memory as long it will be eventually deleted by the garbage collector and isn't query-able anymore.
Event-sourcing solves this. And with how cheap storage is, it should be more prevalent in the industry. IMO the biggest thing holding it back is that there isn't a framework that's plug-and-play (say like Next.js is to React) that provides people with that ability.
I've been working on one in Typescript (with eventual re-writes in other langs. like Rust and Go), but it's difficult even coming up with conventions.
Event sourcing is an expensive solution and I don't mean from a storage perspective — it burns engineering cognitive horsepower quickly on things that don't matter. Do it if you're in finance or whatever. Having been burned by my own "let's event source" impulse on data change tracking systems, I now prefer less sophisticated solutions. Figuring out how to deal with slow projections, watching a projection rebuild go from minutes to hours to a few days as a system I expected to handle a few events/minute go to 20 events/second. Fancy caches can't save you if you want to use that vaunted ability to reconstruct from scratch. Event schema evolution also presents difficult tradeoffs: when old events stop having meaning or evolve in meaning you either end up adding on new event subtypes and variants leaving old cruft to accumulate, or you do migrations and edit history on really large tables.
I'd counsel anyone considering event sourcing to use more "low power" solutions like audit logs or soft deletes (if really necessary) first if possible.
Appreciate your perspective, and it makes me wish there was some kind of online 'engineers learning from their mistakes' forum (rare to see "I burned myself"). To hear hard won knowledge distilled like this is a nice reminder to spend ones complexity budget wisely.
I don't like general advice like this, because it's too general. For many, it's probably good advice. For others, not so much.
Anyone who has worked at a small company selling to large B2B SaaS can attest we get like 20 hits a day on a busy day. Most of that is done by one person in one company, who is probably also the only person from said company you've ever talked to.
From that lens, this is all overkill. It's not bad advice, it's just that it will get quoted for scenarios it doesn't apply. Which also apply to K8S, or microservices at large even, and most 'do as I say' tech blogs.
Shipped and supported enough startup apps to learn this the hard way: users will delete things they shouldn’t, and you will be asked to explain or undo it. Soft deletes and basic metadata (created_at, deleted_by, etc.) have saved me multiple times — not for some future feature, just for basic operational sanity.
Agree, although the acronym in the article could be interpreted to mean “you are going to read it, so index it appropriately”, which is sort of bad advice and can lead to overindexing. There is probably something better for “add appropriate and conventional metadata” (the author suggests updated_at, created_at etc)
Not a huge fan of the example of soft delete, i think hard deletes with archive tables (no foreign key enforcement) is a much much better pattern. Takes away from the main point of the article a bit, but glad the author hinted at deleted_at only being used for soft deletes.
Author is very kind! In practice, many times I saw only the CR/CRU of CRUD getting implemented.
For example: as a company aspires to launch its product, one of the first features implemented in any system is to add a new user. But when the day comes when a customer leaves, suddenly you discover no one implemented off-boarding and cleanup of any sort.
curious that both YAGNI and YAGRI arguments could realistically be made for the same fields. guess it boils down to whether someone’s YAGRI is stronger than their colleague’s YAGNI ( :
This is good advice except for deleted_at. Soft deletion is rarely smart. Deleted things just accumulate and every time you query that table is a new opportunity to forget to omit deleted things. Query performance suffers a lot. It's just a needless complexity.
Instead, just for the tables where you want to support soft delete, copy the data somewhere else. Make a table like `deleteds (tablename text not null, data jsonb not null default '{}')` that you can stuff a serialized copy of the rows you delete from other tables (but just the ones you think you want to support soft delete on).
The theory here is: You don't actually want soft delete, you are just being paranoid and you will never go undelete anything. If you actually do want to undelete stuff, you'll end up building a whole feature around it to expose that to the user anyway so that is when you need to actually think through building the feature. In the meantime you can sleep at night, safe in the knowledge that the data you will never go look at anyway is safe in some table that doesn't cause increased runtime cost and development complexity.
I have a different way of thinking about this: data loss. If you are throwing away data about who performed a delete it is a data loss situation. You should think about whether that’s OK. It probably isn’t.
Well in the same vain that we discuss "points" and talk about the merits, its useful to discuss and understand their counter points. I for one did not know about this and thought it was insightful when building a product that hasn't fully been scoped out and is more greenfield
I don't get why all of the big RDBMSes (PostgreSQL, MariaDB/MySQL, SQL Server, Oracle, ...) don't seem to have built in support for soft deletes up front and center?
Where the regular DELETE wouldn't get rid of the data for real but rather you could query the deleted records as well, probably have timestamps for everything as a built in low level feature, vs having to handle this with a bunch of ORMs and having to remember to put AND deleted_at IS NULL in all of your custom views.
If we like to talk about in-database processing so much, why don't we just put the actual common features in the DB, so that toggling them on or off doesn't take a bunch of code changes in app, or that you'd even be able to add soft deletes to any legacy app that knows nothing of the concept, on a per table basis or whatever.
Maybe my intuition is wrong, but to me this sounds like a violation of the principle of least power.
"Soft deletes" is just a name for a regular write operation, with specific semantics.
Adding a layer of magic to the DB for this doesn't seem right to me.
And applications could have many different requirements for soft deletes, like the article points out. For example, the simplest version would be just a boolean "deleted" column, but it could also be "deleted_at", "deleted_by", etc.
All of these cases require an bunch of code changes anyway, and the more complex ones could interfere with an implementation of this feature at the database level: such a transparent implementation couldn't access app-specific concerns such as user data, for example.
Adding soft deletes to a legacy app that knows nothing about it would only work for a boolean flag and a maybe date-time value, unless the DBMS would also offer triggers for soft deletes etc?
Seems to me to that this capability would make a DBMS much more complicated.
For `updated_at` and `deleted_at` making them nullable and null until touched is incredibly useful.
Answering queries like how many of these were never updated? Or how many of these were never cancelled?
These are not decisions that should be taken solely by whoever is programming the backend.
They need to be surfaced to the product owner to decide. There may very well be reasons pieces of data should not be stored. And all of this adds complexity, more things to go wrong.
If the product owner wants to start tracking every change and by who, that can completely change your database requirements.
So have that conversation properly. Then decide it's either not worth it and don't add any of these "extra" fields you "might" need, or decide it is and fully spec it out and how much additional time and effort it will be to do it as a proper feature. But don't do it as some half-built just-in-case "favor" to a future programmer who may very well have to rip it out.
On a personal project, do whatever you want. But on something professional, this stuff needs to be specced out and accounted for. This isn't a programming decision, it's a product decision.
This entirely depends on the company culture. I worked in teams where every small decision is in the hand of the PO and I've worked in teams where a software engineer is a respected professional enabled to make their own technical decisions. I found the second option to create higher quality software faster.
Also not sure what you mean by additional effort? Created_at, updated_at or soft-deletes are part of most proper frameworks. In Spring all you need is an annotation, I've been using those in major projects and implementation cost is around a few seconds with so far zero seconds of maintenance effort in years of development. At least those fields are solved problems.
even better, assume soft deletes and inform them this is how delete works - if there's a hard requirement for hard delete, they will tell you.
the hypothetical future programmer is you in two weeks.
Some things are trivial and nearly free - created_at, updated_at. I don't think engineers need to bring trivialities like this to a "product owner". Own your craft.
I never worked at a place with product owners, but their post made me appreciate my roles where I'm trusted to help design the product myself. Yeesh.
Being unable to even call the shot of whether a database table should have an updated_at or soft-delete sounds like a Dilbertian hellscape to me.
When the product you're developing is governed by regulations and standards you need to comply, owning your craft is doing things by the book, not adding fields on your own because it might be useful later.
Although those can be more complicated, and it should be clear what they're for and why they exist. Will this result in an object having an updated_at timestamp elsewhere in a larger application? Is it clear which properties that refers to?
I think the tricky part lies on knowing which things can be done without consulting any product owner. I agree that created_at and updated_at don’t cause any harm. deleted_at on the other hand cannot be decided by engineers only (mainly because of GDPR reasons: if something is expected to be totally deleted, then that must be it). As usual, these kind of things are obvious to engineers with years of experience , not so much to newcomers.
A soft delete might not be, for compliance reasons (GDPR and the like). Otherwise I agree.
> don't do it as some half-built just-in-case "favor" to a future programmer who may very well have to rip it out.
in other words - YAGNI !
A product owner may (probably does not) have these things on their radar, it's up to a developer to inform them of industry best practices like these.
Yes. Adding fields to a DB is not a risk-free thing a programmer should just do.
> And all of this adds complexity, more things to go wrong
That's a little vague given this specific example, which appears to be about maintaining some form of informative logging; though I don't think it necessarily needs to be in the form of an DB table.
While I like the YAGRI principle very much, I find that adding
- updated_at
- deleted_at (soft deletes)
- created_by etc
- permission used during CRUD
to every table is a solution weaker than having a separate audit log table.
I feel that mixing audit fields with transactional data in the same table is a violation of the separation of concerns principle.
In the proposed solution, updated_at only captures the last change only. A problem that a separate audit log table is not affected to.
I kinda agree, but don’t underestimate the power of having things where people are looking.
Put your documentation in doc strings where the function is defined - don’t have a separate file in a separate folder for that. It might separate concerns, but no one is looking there.
Similarly if those fields aren’t nullable, someone trying to add new rows will have to fill in something for those metadata fields - and that something will now very likely be what’s needed, rather than not pushing anything to the audit table.
Obviously your app can outgrow these simple columns, but you’re getting value now.
An audit log table often takes a huge amount of space compared to simple fields on the records so there are tradeoffs. Which solution is best depends on how important change logs are.
Event sourcing also works great. You don't need an audit log per se if you already track a history of all commands that introduced changes to your system.
Event sourcing and "the right to be forgotten" are not always easy to marry.
2 replies →
Yep. But Event Sourcing comes with its own set of other problems. Also, I don't think this would apply to OP's post: with Event Sourcing you would not even have those DB tables.
1 reply →
*_at and *_by fields in SQL are just denormalization + pruning patterns consolidated, right?
Do the long walk:
Make the schema fully auditable (one record per edit) and the tables normalized (it will feel weird). Then suffer with it, discover that normalization leads to performance decrease.
Then discover that pruned auditing records is a good middle ground. Just the last edit and by whom is often enough (ominous foreshadowing).
Fail miserably by discovering that a single missing auditing record can cost a lot.
Blame database engines for making you choose. Adopt an experimental database with full auditing history. Maybe do incremental backups. Maybe both, since you have grown paranoid by now.
Discover that it is not enough again. Find that no silver bullet exists for auditing.
Now you can make a conscious choice about it. Then you won't need acronyms to remember stuff!
Another option is audit info could go to another table or datastore entirely.
If you never use it, that data can be dumped to s3 glacier periodically (e.g. after 90 days).
By losing the foreign key you gain flexibility in what you audit. Maybe audit the operation and not the 20 writes it causes.
Fair enough, but now your application is relying on 100% uptime of AWS and S3 and no network failures in between. And what happens if your transaction goes through, but the request to AWS doesn’t? What happens if another operation mutates the target meanwhile before you can retry with current state? Your app is also slowing down since it needs to send the events to S3 and guarantee they got there. Now you are reinventing two-stage commits. Unless you aren’t actually making an audit log and don’t care if events are guaranteed to be logged?
So like OP said, no silver bullets exist for auditing.
1 reply →
My current state is have the database be the current state and use logical replication (CDC) to keep the log of changes in case you need it
It is interesting thinking about record changes as a spectrum towards application logs. At some point too much detail is expensive to store, and you must adopt an archival strategy.
1 reply →
If you see it from the pure SQL point of view, you are in the "blame database engines and adopt an experimental solution".
It is the point where you give up modeling the audit as part of the systems tables.
The drawbacks of this choice are often related to retrieval. It depends on the engine.
I once maintained a system that kept a fully working log replicated instance delayed by 24h, ready for retrieval queries, in addition to regular disk backups (slow costy retrieval).
I am more developer than DBA, so I can probably speak more about modeling solutions than infra-centric solutions.
1 reply →
The problem with this is the audit log is only at the CRUD level which is often too low. Ambiguities can arise. For example if the question is "who published the article" do you look for a create or do you look for an update with published=true? It's even worse when you consider the schema can change over time, so both can be correct but at different points in time. Event sourcing is the way if you want to capture business-level events.
But wait, there's Event Driven Architectures and Event Sourcing, meaning that the events are your log of edits!
Doesn't that also falls on the "blame the database engines and go for an experimental solution"?
I'm not saying databases are blameless. It's just that experiencing the issues they have by yourself is rewarding!
There is also a walk before the long walk of databases. Store things in text files and use basic tools (cat, sed, sh...).
The event driven stuff (like Kafka) reminds me of that. I am not very familiar with it though, just played a little bit with it once or twice.
1 reply →
Additionally, mutable fields will quite often benefit from having a separate edit table which records the old value, the new value, who changed it, and when. Your main table’s created and updated times can be a function of (or a complement to) the edit table.
It is tempting to supernormalize everything into the relations object(id, type) and edit(time, actor_id, object_id, key, value). This is getting dangerously and excitingly close to a graph database implemented in a relational database! Implement one at your peril — what you gain in schemaless freedom you also lose in terms of having the underlying database engine no longer enforcing consistency on your behalf.
> This is getting dangerously and excitingly close to a graph database implemented in a relational database!
This feels like a great unresolved tension in database / backend design - or maybe I'm just not sophisticated enough to notice the solutions?
Is the solution event sourcing and using the relational database as a "read model" only? Is that where the truly sophisticated application developers are at? Is it really overkill for everybody not working in finance? Or is there just not a framework that's made it super easy yet?
Users demand flexible schemas - should we tell them no?
> Additionally, mutable fields will quite often benefit from having a separate edit table which records the old value, the new value, who changed it, and when.
Aren't you describing a non-functional approach to event sourcing? I mean, if the whole point of your system is to track events that caused changes, why isn't your system built around handling events that cause changes?
> supernormalize everything into the relations object(id, type) and edit(time, actor_id, object_id, key, value)
I frankly hate this sort of thing whenever I see it. Software engineers have a tendency to optimize for the wrong things.
Generic relations reduce the number of tables in the database. But who cares about the number of tables in the database? Are we paying per table? Optimize for the data model actually being understandable and consistently enforced (+ bonus points for ease of querying).
One thing I do quite frequently which is related to this (and possibly is a pattern in rails) is to use times in place of Booleans.
So is_deleted would contain a timestamp to represent the deleted_at time for example. This means you can store more information for a small marginal cost. It helps that rails will automatically let you use it as a Boolean and will interpret a timestamp as true.
I consider booleans a code smell. It's not a bug, but it's a suggestion that I'm considering something wrong. I will probably want to replace it with something more meaningful in the future. It might be an enum, a subclass, a timestamp, refactoring, or millions of other things, but the Boolean was probably the wrong thing to do even if I don't know it yet.
The way I think about it: a boolean is usually an answer to a question about the state, not the state itself.
A light switch doesn't have an atomic state, it has a range of motion. The answer to the question "is the switch on?" is a boolean answer to a question whose input state is a range (e.g. is distance between contacts <= epsilon).
This seems at first like a controversial idea, but the more I think about it the more I like this thought technology. Merely the idea of asking myself if there's a better way to store a fact like that will potentially improve designs.
1 reply →
Booleans also force the true/false framing.
E.g. a field called userCannotLoginWithoutOTP.
Then in code "if not userCannotLoginWithoutOTP or otpPresent then..."
Thus may seem easy until you have a few flags to combine and check.
An enum called LoginRequirements with values Password, PasswordAndOTP is one less negation and easier to read.
1 reply →
Tangential: I was recently wishing that bitwise flags had better support in Postgres. For now, bools are just easier to work with
This is all well and good until you need to represent something that happened on Jan 1 1970 00:00 UTC.
32-bit UNIX timestamps are often signed so you can actually go before that, but most UNIX timestamps are 64-bit now, which can represent quite a larger range. And SQL datetime types might have a totally different range.
Not that it really matters; deleted_at times for your database records will rarely predate the existence of said database.
3 replies →
Leave it null for non-deleted items.
« Anytime you store Boolean, a kitten dies » Nobody has ever said that but nobody wants any kitten to die so nobody has ever challenged me anytime I use that statement.
This one little change alone can bring such huge benefits later.
A little while back, I had a conversation with a colleague about sorting entries by "updated at" in the user interface, and to my surprise this was not added by the backend team.
Many of these "we are going to need it"s come from experience. For example in the context of data structures (DS), I have made many "mistakes" that I do correctly a second time. These mistakes made writing algorithms for the DS harder, or made the DS have bad performance.
Sadly, it's hard to transfer this underlying breadth of knowledge and intuition for making good tradeoffs. As such, a one-off tip like this is limited in its usefulness.
Database schemas being perfect out-of-the gate was replaced by reliable migrations.
If it's not data that's essential to serving the current functionality, just add a column later. `updated_at` doesn't have to be accurate for your entire dataset; just set it to `NOW()` when you run the migration.
Sure, migrations are bearable (especially ones that only add columns).
But for the example of the "updated_at" column, or "soft delete" functionality, you only find out you need it because the operations team suddenly discovered they needed that functionality on existing production rows because something weird happened.
In C#-land, we just have it as a standard that ~every table inherits from `ITrackable`, and we wrote a little EF plugin to automatically update the appropriate columns.
public interface ITrackable { DateTime CreatedOn {get; set;} DateTime ModifiedOn {get; set;} }
Saves so much time and hassle.
“Reliable migrations” almost seems like an oxymoron. Migrations are complicated, difficult and error prone. I think there’s a good takeaway here around good initial schema design practices. The less you have to morph your schema overtime, the less of those risky migrations need to run.
2 replies →
Still depends on what the data represent: you could get yourself in a storm of phone calls from customers if after your latest release there's now a weird note saying their saved document was last updated today.
"HOW DARE YOU MODIFY MY DOCUMENTS WITHOUT MY..."
Somewhat related, but I suggest having both the record updated at, and some kind of "user editing updated at". As I've encountered issues where some data migration ends up touching records and bumping the updated at, which shocks users since they see the UI reshuffle and think they have been hacked when they see the records updated at a time they didn't update them.
I mean this is what audit logs are for I'd say: generally speaking you want to know what was changed, by who and why.
So really you probably just want a reference to the tip of the audit log chain.
The perils of UI design wagging the horse.
I like the heuristics described here. However if these things aren't making it into a product spec where appropriate, then I smell some dysfunction that goes beyond what's being stored by default.
Product need (expressed as spec, design, etc) should highlight the failure cases where we would expect fields like these to be surfaced.
I'd hope that any given buisness shouldn't need someone with production database access on hand to inform as to why/when/how 'thing' was deleted. Really we'd want the user (be it 'boss' or someone else) to be able to access that information in a controlled manner.
"What information do we need when something goes wrong?". Ask it. Drill it. Ask it again.
That said, if you can't get those things, this seems a fine way to be pragmatic.
"Wagging the horse" is a great turn of phrase, better than "putting the cart before the dog."
Yes! Why something happened is incredibly important. Gitlab made this mistake hard. We have a medium sized instance with some complex CI pipelines and often they'll just get cancelled and it doesn't say why or even who by. And anyone can do it! The only option is to ask the entire company "did anyone cancel this?"
Counter point: "Soft Deletion Probably Isn't Worth It" https://brandur.org/soft-deletion
But auditing probably is.
I agree with this as written, as think it's important to have some degree of forethought when building out the DB to plan for future growth and needs.
That said, the monkey paw of this would be someone reading it and deciding they should capture and save all possible user data, "just in case", which becomes a liability.
Just curious, how do people feel about this general style of soft deletes currently? Do people still use these in production or prefer to just delete fully or alternatively move deleted rows to a separate tables / schema?
I find the complexity to still feel awkward enough that makes me wonder if deleted_at is worth it. Maybe there are better patterns out there to make this cleaner like triggers to prevent deletion, something else?
As for the article, I couldn't agree more on having timestamps / user ids on all actions. I'd even suggest updated_by to add to the list.
Financial world: records have a "close" or "expire" date which is then purged after some period of time. A deletion doesn't just happen, the record is updated to be "closed" or "expired" and some time after that it's deleted.
Something like a loan could live in a production environment for well over a year after closing, while an internal note may last just a month.
I think soft deletes using timestamptz are a good thing.
Deleting rows directly could mean you're breaking references. For example, say you have a product that the seller wants to delete. Well, what happens if customers have purchased that product? You still want it in the database, and you still want to fulfill the orders placed.
Your backend can selectively query for products, filter out deleted_at for any customer facing queries, but show all products when looking at purchase history.
There are times when deleting rows makes sense, but that's usually because you have a write-heavy table that needs clearing. Yes, soft-deletes requires being careful with WHERE statements filtering out deleted rows, but that's a feature not a bug.
> what happens if customers have purchased that product? You still want it in the database, and you still want to fulfill the orders placed.
You might still want to show to those customers their purchase history including what they bought 25 years ago. For example, my ISP doesn't have anymore that 10 Mb/s fiber optic product I bought im 2000, because it was superseded by 100 Mb/s products and then by 1 Gb/s ones. It's also not my ISP anymore but I use it for the SIM in my phone. That also accumulated a number of product changes along the years.
And think about the inventory of eshops with a zillion products and the archive of the pady orders. Maybe they keep the last few years, maybe everything until the db gets too large.
There can be legal requirements to retain data for a specified time for law enforcement and audits, while at the same time other legal requirements have you requiring to delete data upon customer request.
Doing this with pure 'hard' deletes is not possible, unless you maintain 2 different tables, one of which would still have the soft delete explicit or implicit. You could argue the full db log would contain the data for the former requirement, but while academicly correct this does not fly in practice.
> Maybe there are better patterns out there to make this cleaner
SQL:2011 temporal tables are worth a look.
If you have a good audit log, it really doesn't matter. You can always restore it if need be.
If you have no audit log(or a bad one), like lots of apps, then you have to care a lot.
Personally, I just implement a good audit log and then I just delete with impunity. Worst case scenario, someone(maybe even me) made a mistake and I have to run undo_log_audit() with the id of the audit log entry I want to put back. Nearly zero hassle.
The upside, when something goes wrong, I can tell you who, what and when. I usually have to infer the why, or go ask a human, but it's not usually even difficult to do that.
Can you share more about what makes a good audit log? My company doesn’t currently have one and I’m a little lost on where to start.
Should this be at the application code level, or the ORM, or the database itself?
3 replies →
Always soft-deletion first. Then it gets exported to a separate archive and only then, after some time and may be attempted to be fully deleted from the initial base.
As an acronym, it's easy to be misremembered as "You ARENT gonna read it" (based on the popularity of yagni) - and have the opposite advice spread..
How about: IIPTWNIRTIoSoDIQMBSNRTL
(It Is Probable That While Not Immediately Required The Implementation of Storage of Data In Question May Be Simpler Now Rather Than Later)
I've gone ahead and included additional detail in the acronym in the event that the clarity is required later, as this would be difficult to retrofit into a shorter, more-established acronym.
Why do I feel like that should be a cheat code in a FPS game?
YRIS: you'll read it someday
It's a terrible post. What it suggests is to turn your head off and follow overly generalised principle. I guess when somebody invent yet another acronym it is a red flag.
Data has its own life cycles in every area it passes through. And it's part of requirements gathering to find those cycles: the dependent systems, the teams, and the questions you need to answer. Mindlessly adding fields won't save you in every situation.
Bonus point: when you start collecting questions while designing your service, you'll discover how mature your colleagues' thinking is.
Five years ago everybody would lough about "soft deletes" or "marked as deleted". Whoever thought this is a good idea from a data protection perspective? You also lying in the face of your users with such a behavior. Shame.
Except almost every database (and most storage devices nowadays) works this way - mark an entry as deleted, then batch delete a lot of entries during garbage collection. It's fundamentally impossible to efficiently erase a record from the middle of a file, except maybe by using an encryption tree, which would still be fairly inefficient.
Actually erasing data is quite hard. Soft deletes doesn't add any new lies, they just move the lies to the upper layer.
Come on. With a manual "marked as deleted" it stays as this forever, it is not deleted and never will, and the "deleted" data lands also in database backups, is still query-able and so on. I do not care if the deleted data stays for a while on disk or in memory as long it will be eventually deleted by the garbage collector and isn't query-able anymore.
Event-sourcing solves this. And with how cheap storage is, it should be more prevalent in the industry. IMO the biggest thing holding it back is that there isn't a framework that's plug-and-play (say like Next.js is to React) that provides people with that ability.
I've been working on one in Typescript (with eventual re-writes in other langs. like Rust and Go), but it's difficult even coming up with conventions.
Event sourcing is an expensive solution and I don't mean from a storage perspective — it burns engineering cognitive horsepower quickly on things that don't matter. Do it if you're in finance or whatever. Having been burned by my own "let's event source" impulse on data change tracking systems, I now prefer less sophisticated solutions. Figuring out how to deal with slow projections, watching a projection rebuild go from minutes to hours to a few days as a system I expected to handle a few events/minute go to 20 events/second. Fancy caches can't save you if you want to use that vaunted ability to reconstruct from scratch. Event schema evolution also presents difficult tradeoffs: when old events stop having meaning or evolve in meaning you either end up adding on new event subtypes and variants leaving old cruft to accumulate, or you do migrations and edit history on really large tables.
I'd counsel anyone considering event sourcing to use more "low power" solutions like audit logs or soft deletes (if really necessary) first if possible.
Appreciate your perspective, and it makes me wish there was some kind of online 'engineers learning from their mistakes' forum (rare to see "I burned myself"). To hear hard won knowledge distilled like this is a nice reminder to spend ones complexity budget wisely.
Care to share what you've been working on?
I don't like general advice like this, because it's too general. For many, it's probably good advice. For others, not so much.
Anyone who has worked at a small company selling to large B2B SaaS can attest we get like 20 hits a day on a busy day. Most of that is done by one person in one company, who is probably also the only person from said company you've ever talked to.
From that lens, this is all overkill. It's not bad advice, it's just that it will get quoted for scenarios it doesn't apply. Which also apply to K8S, or microservices at large even, and most 'do as I say' tech blogs.
>I don't like general advice like this, because it's too general. For many, it's probably good advice. For others, not so much.
That's true for any other good advice you may have heard of.
Shipped and supported enough startup apps to learn this the hard way: users will delete things they shouldn’t, and you will be asked to explain or undo it. Soft deletes and basic metadata (created_at, deleted_by, etc.) have saved me multiple times — not for some future feature, just for basic operational sanity.
> But I've never heard someone complain about a table having too many timestamps.
I do. Each one is 8 bytes. At the billions of rows scale, that adds up. Disk is cheap, but not free; more importantly, memory is not cheap at all.
Agree, although the acronym in the article could be interpreted to mean “you are going to read it, so index it appropriately”, which is sort of bad advice and can lead to overindexing. There is probably something better for “add appropriate and conventional metadata” (the author suggests updated_at, created_at etc)
Not a huge fan of the example of soft delete, i think hard deletes with archive tables (no foreign key enforcement) is a much much better pattern. Takes away from the main point of the article a bit, but glad the author hinted at deleted_at only being used for soft deletes.
Author is very kind! In practice, many times I saw only the CR/CRU of CRUD getting implemented.
For example: as a company aspires to launch its product, one of the first features implemented in any system is to add a new user. But when the day comes when a customer leaves, suddenly you discover no one implemented off-boarding and cleanup of any sort.
Related from a few years ago: PAGNIs, for Probably Are Gonna Need Its - https://simonwillison.net/2021/Jul/1/pagnis/
OT As the great Alone ("Last Psychiatrist") said, "if you read it, it's for you" - IYRIIFY
Literally just have a good audit log and then you get all of this for free and more.
curious that both YAGNI and YAGRI arguments could realistically be made for the same fields. guess it boils down to whether someone’s YAGRI is stronger than their colleague’s YAGNI ( :
This is good advice except for deleted_at. Soft deletion is rarely smart. Deleted things just accumulate and every time you query that table is a new opportunity to forget to omit deleted things. Query performance suffers a lot. It's just a needless complexity.
Instead, just for the tables where you want to support soft delete, copy the data somewhere else. Make a table like `deleteds (tablename text not null, data jsonb not null default '{}')` that you can stuff a serialized copy of the rows you delete from other tables (but just the ones you think you want to support soft delete on).
The theory here is: You don't actually want soft delete, you are just being paranoid and you will never go undelete anything. If you actually do want to undelete stuff, you'll end up building a whole feature around it to expose that to the user anyway so that is when you need to actually think through building the feature. In the meantime you can sleep at night, safe in the knowledge that the data you will never go look at anyway is safe in some table that doesn't cause increased runtime cost and development complexity.
How do you distinguish from "you aren't gonna read it"? The acronym is poorly designed.
YAGRI proponents organized themselves into a community to develop their...YAGRIculture.
I'll show myself out.
I have a different way of thinking about this: data loss. If you are throwing away data about who performed a delete it is a data loss situation. You should think about whether that’s OK. It probably isn’t.
To every point there is a counterpoint. So what?
Well in the same vain that we discuss "points" and talk about the merits, its useful to discuss and understand their counter points. I for one did not know about this and thought it was insightful when building a product that hasn't fully been scoped out and is more greenfield
[dead]