When a mistake is found¶
A binding can break, and it breaks at the earliest phase that can see it. Checking sees the configuration; evaluation sees the systems:
| What is wrong | Found | What is said |
|---|---|---|
tap("order_event") |
checking | check error[unknown-name]: no endpoint order_event is configured; did you mean order_events? |
source("order_events") |
checking | check error[type-mismatch]: order_events is a kafka endpoint, read by tap as a Stream, not by source as a Seq |
a stream written by target("orders") |
checking | check error[type-mismatch]: target("orders") writes a Table or a sequence of a declared record, not Stream<Event> |
sink("ledger") |
checking | check error[type-mismatch]: endpoint ledger is a sqlite endpoint, which sink does not write; write it whole with target |
a password in .hql/config.toml |
loading the vault | url holds credentials; name the environment variable that holds the address, url_env = "VARIABLE" |
SHOP_DSN not set |
evaluation, before the first row | execute failure[io-failure]: endpoint orders: the environment variable SHOP_DSN, which dsn_env names, is not set |
a raw order read as Event |
evaluation, at that order | execute failure[io-failure]: endpoint raw_events, field amount: "free" is not a decimal number |
| a write to a wildcard subject | evaluation, at the write | execute failure[io-failure]: subject orders.> holds a wildcard, which reads many subjects and writes none; a message is published to one subject |
A secret is named by the variable that holds it and never printed. A
failure at evaluation names the endpoint by its id and the value by its
field; what it never does is continue past a value it could not read and
count it later as missing. A pipeline that must continue says so in its
shape, as day.hql does with validate and fork: the order that would
have stopped it is a row in the dead-letter subject instead.