Pin smoke test to the 0.2.0-stable API surface

The smoke test's job is "downstream Ant/Ivy consumers can pick up
the published artefact" — that goal is met by exercising the
parser, types, and bounds methods that have been on Maven Central
since the first release.

Drop the EdtfFormatter call from Smoke.java because that class only
ships in 0.3.0+; with the previous wiring the daily workflow would
fail to compile against the currently-published 0.2.0 jar with
"package io.github.openhistoricalmap.edtf.format does not exist".
The interval upper-endpoint type is now exercised instead.

build.xml's substring assertions are tightened: instead of looking
for "May 2020" (formatter output) they now look for the canonical
"2020-05-15", the EdtfType "DATE", and the "2020/.." round-trip,
all of which the 0.2.0 API produces.

README.md adds a sentence explaining the version-tracking choice.

Co-Authored-By: Claude <noreply@anthropic.com>

Author: jeffreyameyer

smoke/README.md

@@ -19,10 +19,14 @@ that:
    `io.github.openhistoricalmap:edtf:${edtf.version}` from Maven
    Central into `lib/`.
 2. `javac` compiles `src/Smoke.java`, which calls
-   `Edtf.parse("2020-05-15")` and prints the round-tripped EDTF
-   string and the lower bound.
-3. `java` runs the compiled class and the script asserts the
-   output matches expected values; any deviation exits non-zero.
+   `Edtf.parse("2020-05-15")` and a `2020/..` open-upper interval.
+3. `java` runs the compiled class and the script asserts on the
+   captured stdout; any deviation exits non-zero.
+
+The smoke source uses only API that has been stable since 0.2.0
+(the first published release). The daily workflow therefore
+exercises whichever version is currently on Maven Central without
+needing to be in lockstep with new feature releases.
 
 ## Prerequisites
 

smoke/build.xml

@@ -56,8 +56,9 @@
 
         <condition property="smoke.passed">
             <and>
-                <contains string="${smoke.stdout}" substring="2020-05-15"/>
-                <contains string="${smoke.stdout}" substring="May 2020"/>
+                <contains string="${smoke.stdout}" substring="canonical: 2020-05-15"/>
+                <contains string="${smoke.stdout}" substring="type: DATE"/>
+                <contains string="${smoke.stdout}" substring="interval canonical: 2020/.."/>
             </and>
         </condition>
         <fail unless="smoke.passed"

smoke/src/Smoke.java

@@ -1,30 +1,28 @@
 import io.github.openhistoricalmap.edtf.Edtf;
 import io.github.openhistoricalmap.edtf.EdtfTemporal;
-import io.github.openhistoricalmap.edtf.format.EdtfFormatter;
-import java.util.Locale;
 
 /**
  * Smoke test asserting that the published edtf jar is usable from
- * a vanilla Ant + Ivy classpath. Prints two lines, both of which are
- * checked by build.xml's `test` target:
- *   1. The canonical round-tripped EDTF string.
- *   2. The English-localised formatter output.
+ * a vanilla Ant + Ivy classpath. Uses only the API that has been
+ * stable since the first published release (0.2.0), so the daily
+ * smoke workflow exercises whichever version is on Maven Central
+ * without coupling to features that may not yet be released.
+ *
+ * Prints lines that build.xml's `test` target checks for substrings.
  */
 public class Smoke {
     public static void main(String[] args) {
         String input = "2020-05-15";
         EdtfTemporal t = Edtf.parse(input);
-
         System.out.println("canonical: " + t.toEdtfString());
-        System.out.println("localised: " + EdtfFormatter.forLocale(Locale.US).format(t));
+        System.out.println("type: " + t.type());
+        System.out.println("level: " + t.level());
         System.out.println("min: " + t.min());
         System.out.println("max: " + t.max());
-        System.out.println("type: " + t.type());
 
-        // A second exercise: an open-ended interval with the formatter.
+        // Cover an interval too.
         EdtfTemporal i = Edtf.parse("2020/..");
         System.out.println("interval canonical: " + i.toEdtfString());
-        System.out.println("interval localised: "
-            + EdtfFormatter.forLocale(Locale.US).format(i));
+        System.out.println("interval upper-open: " + i.upper().getClass().getSimpleName());
     }
 }