Base Mechanism to Concrete Realization Cheat Sheet #

Use this note after identifying:

normalized path
write shape
base mechanism

This is the next step: grounding the mechanism in concrete components.

1. Insert / Create #

Used for:

create post
create order
create comment
create file

Typical realization:

primary OLTP store
- Postgres / MySQL / DynamoDB / Cassandra row insert
optional idempotency key table
optional event emitted for downstream projections

Mental model:

mostly store-centered
plus optional async fanout

2. Versioned Overwrite #

Used for:

edit profile
rename file
update document title
update feature flag

Typical realization:

primary OLTP store
version column / ETag / update timestamp guard
optional cache invalidation
optional projection refresh

Mental model:

mostly store-centered
often plus cache and projection updates

3. Guarded Conditional Update #

Used for:

follow/unfollow
revoke share
cancel booking
approve refund
accept suggestion
restore file

Typical realization:

primary OLTP store with:
- conditional update
- WHERE version = ?
- or explicit state predicate
optional transaction if multiple rows are affected
optional async workers updating counters, search, feed, or audit

Mental model:

source-truth store
plus async secondary systems

4. Append Log / Append Table #

Used for:

comments
audit log
activity events
bids
metrics/spans

Typical realization:

append-only DB table
or Kafka / Pulsar / event log
or time-series store
optional projection/index builders

Mental model:

append store for truth
downstream consumers for views and indexes

5. Materialized View #

Used for:

home feed
search results
leaderboard
top K
dashboard
hashtag search

Typical realization:

source-truth store
change stream / outbox / event queue
async worker(s)
read-model store:
- feed table
- Redis sorted set
- Elasticsearch/OpenSearch index
- precomputed aggregation table

Mental model:

almost always multi-component

6. Counter / Projection Update #

Used for:

follower count
like count
top-tag stats
aggregate ratings

Typical realization:

source-truth store
async delta consumer or synchronous write-through
counter table / Redis / aggregate row
periodic recompute or repair job

Mental model:

store + worker
sometimes cache layer

7. Claim / Lease #

Used for:

worker claims job
lock acquire
worker claims reminder delivery
executor claims action

Typical realization:

primary coordination store or source-truth row
expiry timestamp
claim token / epoch / fencing token
heartbeat / renew path
timeout worker to reclaim expired claims

Mental model:

store + timeout worker + token semantics

8. Guarded Update + Hold / Lease #

Used for:

seat hold
room hold
appointment slot hold

Typical realization:

inventory/resource store
hold record or held-state field
conditional write / transaction
expiry scheduler/worker
booking workflow object on confirmation

Mental model:

resource store + hold store/state + expiry worker

9. Guarded Frontier Update #

Used for:

campaign recipient frontier
crawler cursor
scheduler next-fire cursor
batch progress marker

Typical realization:

frontier row/document in store
worker or scheduler claims range
conditional version update on frontier
batch/run records optionally created

Mental model:

store + scheduler/worker loop

10. Due Scan + Create Target #

Used for:

reminders becoming due
delayed tasks becoming runnable
retries after backoff
scheduled notifications

Typical realization:

scheduled object table indexed by run_at
scheduler scan loop
create runnable job in queue/table
worker claim/lease path after creation

Mental model:

store + due-time index + scheduler + queue/worker

11. Versioned Snapshot Publication #

Used for:

feature flags
policy engine
config delivery
API key validator snapshot

Typical realization:

control-plane source store
snapshot builder / recompute job
distribution channel:
- watch stream
- long poll
- push channel
local in-memory cache on evaluators/consumers

Mental model:

control-plane store + publisher + local cache

12. Search Index #

Used for:

document search
business search
channel search
skill search

Typical realization:

source-truth store
indexing pipeline / outbox consumer
Elasticsearch / OpenSearch / Typesense / similar
query frontend

Mental model:

source store + indexing worker + search engine

13. Ranking / Top K #

Used for:

trending hashtags
top creators
weekly top performers
top restaurants

Typical realization:

source-truth store
aggregate stats store
ranking materializer
heap / sorted set / ranked table
optional Redis cache

Mental model:

source store + stats pipeline + ranked projection

14. Workflow Orchestration #

Used for:

refund processing
checkout
booking confirmation
moderation review
campaign sending

Typical realization:

workflow state row/object
maybe local transaction across a few source rows
async job queue for side effects
outbox/event publishing
retry / reconciliation workers

Mental model:

workflow store + async workers + outbox

Quick Grouping #

Mostly store-centered #

insert/create
versioned overwrite
guarded conditional update
source read

Usually store + async workers #

counters
top K
search indexing
feed materialization
audit trails
workflow side effects

Usually store + scheduler/claim logic #

holds with expiry
delayed jobs
reminders
leases/locks
frontier processing

Usually control-plane + local cache #

config
feature flags
policy engines
API key validation snapshots

Interview One-Liner #

After I pick a base mechanism, I ground it in concrete components: source-truth store for correctness, async workers for projections and side effects, schedulers for due-time or expiry logic, search/ranking systems for derived reads, and local caches or snapshots for hot-path evaluation when bounded staleness is acceptable.