@amplytools/react-native-amply-sdk 0.11.0 → 0.12.0

package/AmplyReactNative.podspec

@@ -22,7 +22,7 @@ Pod::Spec.new do |s|
   s.dependency 'React-RCTFabric'
 
   # Amply KMP SDK from CocoaPods
-  s.dependency 'AmplySDK', '~> 0.11.0'
+  s.dependency 'AmplySDK', '~> 0.12.0'
 
   s.libraries = 'c++'
   s.pod_target_xcconfig = {

package/android/build.gradle

@@ -15,7 +15,7 @@ buildscript {
 apply plugin: 'com.android.library'
 apply plugin: 'org.jetbrains.kotlin.android'
 
-def amplySdkVersion = project.findProperty('amplySdkVersion') ?: System.getenv('AMPLY_SDK_VERSION') ?: '0.11.0'
+def amplySdkVersion = project.findProperty('amplySdkVersion') ?: System.getenv('AMPLY_SDK_VERSION') ?: '0.12.0'
 group = 'tools.amply'
 version = amplySdkVersion
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@amplytools/react-native-amply-sdk",
-  "version": "0.11.0",
+  "version": "0.12.0",
   "description": "React Native SDK for Amply: Mobile SDK to orchestrate in-app experiences and campaigns remotely, without app releases",
   "license": "MIT",
   "repository": {
@@ -33,12 +33,13 @@
     "build": "expo-module build",
     "build:plugin": "expo-module build plugin",
     "clean": "expo-module clean",
-    "test": "node scripts/codegen.js --check && expo-module test && yarn check:seam-tokens",
+    "test": "node scripts/codegen.js --check && expo-module test && yarn check:seam-tokens && yarn check:sample-parity",
     "prepare": "expo-module prepare",
     "prepublishOnly": "expo-module prepublishOnly",
     "codegen": "node scripts/codegen.js",
     "codegen:check": "node scripts/codegen.js --check",
-    "check:seam-tokens": "bash scripts/check-seam-tokens.sh"
+    "check:seam-tokens": "bash scripts/check-seam-tokens.sh",
+    "check:sample-parity": "bash scripts/check-sample-parity.sh"
   },
   "peerDependencies": {
     "expo": ">=54.0.0",

package/scripts/check-sample-parity.sh

@@ -0,0 +1,263 @@
+#!/usr/bin/env bash
+#
+# Every control the sample spec requires must exist in every sample that claims the
+# platform, tagged with its manifest id. Nothing else notices when one of them is missing.
+#
+# The six Amply samples are six copies of the same screen, and they drifted because
+# copy-paste has no compiler. In the KMP repo, `ios-spm-sample/HomeViewController.swift`
+# reached 886 lines while its CocoaPods twin sat at 566 — the same screen, forked and then
+# edited on one side only, with `GateDemo.swift` existing in exactly one of them. Both
+# projects built clean the whole time. Nobody noticed, because nothing checked. The two RN
+# examples here forked the same way: `bare/src/hooks/useAmplyDemo.ts` and
+# `expo/src/hooks/useAmplyDemo.ts` are two copies of one hook, and `bare` still imports the
+# SDK under a legacy alias no integrator would type.
+#
+# This is that check. It greps each example's own sources for every manifest id required on
+# the `rn` column and fails naming the id, the example, and where it looked. It is a parity
+# check, not a UI test: it answers "is this control present in this sample", which is the
+# only question consistency actually turns on.
+#
+# Spec:     multiplatform-library-template/samples/SAMPLE_APP_SPEC.md   (canonical, KMP repo)
+# Manifest: example/sample-manifest.txt   (verbatim copy of the KMP repo's canonical file;
+#                                          this script diffs the two when both are present)
+#
+# Usage: bash scripts/check-sample-parity.sh
+set -uo pipefail
+
+cd "$(dirname "$0")/.."
+
+MANIFEST="example/sample-manifest.txt"
+SPEC_PATH="multiplatform-library-template/samples/SAMPLE_APP_SPEC.md"
+
+if [ ! -f "$MANIFEST" ]; then
+  echo "check-sample-parity: FAIL — missing $MANIFEST" >&2
+  exit 1
+fi
+
+# Quote characters an id can be wrapped in: JSX uses ", JS also uses ' and ` for template
+# literals. The id literal must sit next to `testID` on one line.
+Q="[\"'\`]"
+
+# ---------------------------------------------------------------------------------------
+# The canonical copy. The manifest lives in the KMP repo; this repo carries a duplicate
+# because it is published to npm and mirrored publicly and so cannot depend on a sibling
+# checkout existing. Duplication that nothing checks is drift with extra steps, so: when
+# the KMP repo is reachable, diff it and fail on any difference. When it is not, say so
+# loudly and print the revision, so a stale copy shows up in the output instead of passing
+# silently.
+# ---------------------------------------------------------------------------------------
+revision=$(grep -m1 '^# revision:' "$MANIFEST" | sed 's/^# revision:[[:space:]]*//')
+if [ -z "$revision" ]; then
+  echo "check-sample-parity: FAIL — $MANIFEST has no '# revision:' header" >&2
+  exit 1
+fi
+
+peer=""
+for candidate in \
+  "${AMPLY_SAMPLE_MANIFEST_PEER:-}" \
+  "../multiplatform-library-template/samples/sample-manifest.txt"
+do
+  [ -n "$candidate" ] && [ -f "$candidate" ] && { peer="$candidate"; break; }
+done
+
+if [ -n "$peer" ]; then
+  if ! diff -u "$peer" "$MANIFEST" > /tmp/check-sample-parity-peer.diff 2>&1; then
+    echo "check-sample-parity: FAIL — this repo's manifest copy has diverged from the canonical one." >&2
+    echo "                     canonical: $peer" >&2
+    echo "                     copy:      $MANIFEST" >&2
+    sed 's/^/                     /' /tmp/check-sample-parity-peer.diff >&2
+    rm -f /tmp/check-sample-parity-peer.diff
+    exit 1
+  fi
+  rm -f /tmp/check-sample-parity-peer.diff
+  peer_state="verified byte-identical against $peer"
+else
+  peer_state="NOT VERIFIED — the KMP repo was not found; set AMPLY_SAMPLE_MANIFEST_PEER to check it"
+fi
+
+# ---------------------------------------------------------------------------------------
+# Parse the manifest into: id <TAB> android <TAB> ios <TAB> rn <TAB> note
+# ---------------------------------------------------------------------------------------
+rows=$(
+  awk -F'|' '
+    /^[[:space:]]*#/ { next }
+    /^[[:space:]]*$/ { next }
+    NF < 4 { next }
+    {
+      for (i = 1; i <= NF; i++) { gsub(/^[[:space:]]+|[[:space:]]+$/, "", $i) }
+      note = (NF >= 5) ? $5 : ""
+      print $1 "\t" $2 "\t" $3 "\t" $4 "\t" note
+    }
+  ' "$MANIFEST"
+)
+
+if [ -z "$rows" ]; then
+  echo "check-sample-parity: FAIL — parsed zero rows from $MANIFEST; the parser is broken" >&2
+  exit 1
+fi
+
+row_count=$(printf '%s\n' "$rows" | wc -l | tr -d ' ')
+
+# ---------------------------------------------------------------------------------------
+# The examples. Each one searches its own sources plus the shared UI module. An absent root
+# is skipped rather than fatal; an example with NO existing root at all is fatal, because
+# that means this list is stale.
+#
+# The shared module is on BOTH examples' search paths, and that is a trap this check walks
+# into unless it is told not to. A control id living in example/shared satisfies the grep
+# for bare and expo alike — including for a host that never imports the module and still
+# renders its own old screen. That is not hypothetical: on the day the shared module landed,
+# `expo` reported "all 40 required controls present" while app/index.tsx was still building
+# its own ScrollView out of useAmplyDemo. Green, and completely wrong.
+#
+# The native samples do not have this hole. Both Xcode projects list ../ios-shared/Sources
+# and both Gradle modules include the shared srcDir, so on those platforms presence in the
+# shared source IS compilation into the target. Metro resolving a package is not the same
+# claim: a module on the search path is only rendered if something imports it.
+#
+# So each RN host must also prove it MOUNTS the shared module. `mount` below is a file that
+# must exist and must import SHARED_PKG. Without this, the consolidation can be reported
+# complete while half of it was never wired.
+# ---------------------------------------------------------------------------------------
+SHARED_PKG="@amply/sample-shared"
+samples=(
+  "bare|example/bare/src example/shared|example/bare/src/App.tsx"
+  "expo|example/expo/app example/expo/src example/expo/components example/shared|example/expo/app/index.tsx"
+)
+
+ATTR="testID"
+INCLUDES="--include=*.ts --include=*.tsx --include=*.js --include=*.jsx"
+
+fail=0
+checked_total=0
+
+echo "check-sample-parity: manifest revision $revision — $row_count rows"
+echo "                     peer copy: $peer_state"
+echo
+
+for entry in "${samples[@]}"; do
+  name="${entry%%|*}"
+  rest="${entry#*|}"
+  roots="${rest%%|*}"
+  mount="${rest#*|}"
+
+  existing_roots=""
+  for r in $roots; do
+    [ -d "$r" ] && existing_roots="$existing_roots $r"
+  done
+
+  if [ -z "$existing_roots" ]; then
+    echo "check-sample-parity: FAIL — example '$name' has no source root on disk (looked in: $roots)" >&2
+    echo "                     the sample list in this script is stale." >&2
+    fail=1
+    continue
+  fi
+
+  # Does this host actually MOUNT the shared module, or does it merely have it on a search
+  # path? Only the second is true of a host still rendering its own screen, and the id grep
+  # below cannot tell the difference — every shared id would satisfy it either way.
+  if [ -d "example/shared" ]; then
+    if [ ! -f "$mount" ]; then
+      echo "check-sample-parity: FAIL — example '$name' has no entry point at $mount" >&2
+      echo "                     the mount path in this script is stale." >&2
+      fail=1
+      continue
+    fi
+    if ! grep -qF "$SHARED_PKG" "$mount"; then
+      echo "check-sample-parity: FAIL — example '$name' does not mount the shared UI module." >&2
+      echo "                     $mount never imports $SHARED_PKG, so this example still" >&2
+      echo "                     renders its own screen. Every control id below would be" >&2
+      echo "                     satisfied by example/shared regardless — a green that means" >&2
+      echo "                     nothing. Mount the shared module in $mount." >&2
+      fail=1
+      continue
+    fi
+  fi
+
+  missing=()
+  unbound=()
+  required=0
+
+  while IFS=$'\t' read -r id android ios rn note; do
+    [ "$rn" = "x" ] || continue
+    required=$((required + 1))
+
+    id_esc=$(printf '%s' "$id" | sed 's/[.[\*^$()+?{}|\\]/\\&/g')
+
+    case "$note" in
+      launcharg:*)
+        # Not a UI control — the launch key just has to appear in the source. On RN the
+        # transport is a deeplink query param rather than a launch argument (neither
+        # example can read one without a new native dependency); the key spelling is the
+        # same, which is why one manifest row covers all three platforms.
+        # shellcheck disable=SC2086
+        if ! grep -RIlE $INCLUDES -e "$id_esc" $existing_roots > /dev/null 2>&1; then
+          missing+=("$id")
+        fi
+        continue
+        ;;
+      prefix:*) pattern="${Q}${id_esc}\\." ;;
+      *)        pattern="${Q}${id_esc}${Q}" ;;
+    esac
+
+    # shellcheck disable=SC2086
+    hits=$(grep -RInE $INCLUDES -e "$pattern" $existing_roots 2>/dev/null)
+    if [ -z "$hits" ]; then
+      missing+=("$id")
+      continue
+    fi
+    if ! printf '%s\n' "$hits" | grep -q "$ATTR"; then
+      unbound+=("$id")
+    fi
+  done <<< "$rows"
+
+  checked_total=$((checked_total + required))
+
+  if [ ${#missing[@]} -eq 0 ] && [ ${#unbound[@]} -eq 0 ]; then
+    echo "  OK   $name (rn) — all $required required controls present"
+    continue
+  fi
+
+  fail=1
+  echo "  FAIL $name (rn) — $((${#missing[@]} + ${#unbound[@]})) of $required required controls not satisfied"
+  echo "       searched:$existing_roots"
+  if [ ${#missing[@]} -gt 0 ]; then
+    echo "       missing (${#missing[@]}):"
+    for id in "${missing[@]}"; do echo "         - $id"; done
+  fi
+  if [ ${#unbound[@]} -gt 0 ]; then
+    echo "       present but not on a line with \`$ATTR\` (${#unbound[@]}):"
+    for id in "${unbound[@]}"; do echo "         - $id"; done
+  fi
+  echo
+done
+
+# ---------------------------------------------------------------------------------------
+# Retired ids: controls that were required once and must NOT come back. A presence-only
+# check cannot see a control that should have been deleted, so without this a removal is
+# held together by trust. Failed on sight, wherever they appear.
+# ---------------------------------------------------------------------------------------
+retired_ids=$(grep -E '^# +retired: +amply\.' "$MANIFEST" | sed -E 's/^# +retired: +([^ ]+).*/\1/')
+for rid in $retired_ids; do
+  rid_esc=$(printf '%s' "$rid" | sed 's/[.[\*^$()+?{}|\\]/\\&/g')
+  # shellcheck disable=SC2086
+  retired_roots=""
+  for r in example/bare/src example/expo/app example/expo/src example/expo/components example/shared; do [ -d "$r" ] && retired_roots="$retired_roots $r"; done
+  # shellcheck disable=SC2086
+  hits=$(grep -RIlE --include=*.ts --include=*.tsx --include=*.js --include=*.jsx -e "${Q}${rid_esc}${Q}" $retired_roots 2>/dev/null || true)
+  if [ -n "$hits" ]; then
+    echo "check-sample-parity: FAIL — retired control '$rid' is still in the source:" >&2
+    printf '                       %s\n' $hits >&2
+    echo "                     It was deliberately removed; see the manifest's Retired block." >&2
+    fail=1
+  fi
+done
+
+if [ "$fail" -ne 0 ]; then
+  echo "check-sample-parity: FAIL" >&2
+  echo "                     Each missing id is a control $SPEC_PATH requires and the example" >&2
+  echo "                     does not have. Tag the real widget with it: testID=\"<id>\"." >&2
+  exit 1
+fi
+
+echo "check-sample-parity: OK — manifest revision $revision, $checked_total control checks across ${#samples[@]} examples."