Ankur
c438afc33b
37 runnable examples covering the eight feature posts on ankurm.com, verified against Jackson 3.2.1 on Temurin 21.0.5. Every output committed under docs/ was produced by run-all.sh. Also documents 11 places where the published snippets do not compile or do not behave as printed against a real Jackson 3 build - most notably that writeValueAsString(List<Base>) silently drops the polymorphic type discriminator, so the post's serialised output cannot be read back.
2.9 KiB
Part 6 — Streaming API and the tree model
Post: https://ankurm.com/jackson-streaming-api-jsonnode/
Renames that break the post's code verbatim:
| Jackson 2 | Jackson 3 |
|---|---|
com.fasterxml.jackson.core.JsonFactory |
tools.jackson.core.json.JsonFactory (note the extra .json) |
parser.getCurrentName() |
parser.currentName() |
parser.getText() |
parser.getString() |
JsonToken.FIELD_NAME |
JsonToken.PROPERTY_NAME |
gen.writeNumberField / writeStringField |
gen.writeNumberProperty / writeStringProperty |
node.asText() |
node.asString() |
factory.createParser(file) |
factory.createParser(ObjectReadContext.empty(), file) |
The comparison post's performance section gives the import as
tools.jackson.core.JsonFactory, which does not resolve — the class is in the
.json subpackage.
G01 — filtering a large array
ERROR: Database connection failed
Total errors found: 1
G02 — writing a million records
The post's loop, kept at its full 1,000,000 iterations, with the heap delta measured so "constant memory usage" is a number.
records written : 1000000
file size : 30 MB
elapsed : 129 ms
heap delta : 0 MB <- the document is never held in memory
G03 — the tree model
Customer: Alice
KB-01 x2
MS-42 x1
Has discount: false
path(missing) : (class MissingNode)
get(missing) : null
deep path : '<default>'
treeToValue : CustomerRecord[name=Alice, tier=gold]
The path(missing) / get(missing) pair is the whole argument for path(): it returns
a MissingNode that chains safely, where get() returns null and the next call NPEs.
G04 — the decision table, measured
G04ThreeApproachesMeasured.java
The post closes with a table asserting relative memory and verbosity. This runs the same filter three ways over a generated 20 MB file.
input file : 20 MB, 200000 entries
data binding (readValue) errors=4000 524 ms heap delta 47 MB
tree model (readTree) errors=4000 341 ms heap delta 102 MB
streaming (JsonParser) errors=4000 78 ms heap delta 1 MB
Single-shot measurements on a 2-core container, not JMH — treat the ordering as the result, not the absolute figures. Two things do stand out and are stable across runs: the tree model costs roughly twice the heap of data binding for the same document, and streaming is the only approach whose heap does not scale with input size.