Partial changes from update-imports branch

.github/workflows/check-reproducibility.yml

@@ -40,6 +40,8 @@ jobs:
     - name: Clean caches & workspace
       run: make clean
     - run: echo "LOGURU_COLORIZE=NO" >> $GITHUB_ENV
+    - name: Build
+      run: cd src/scripts/merge_tools/merging && ./gradlew -q shadowJar
     - name: make check-merges-reproducibility
       run: |
         git config --global user.email "[email protected]"

Makefile

@@ -134,7 +134,7 @@ clean-local:
 check-merges-reproducibility:
 	@echo "Running replay_merge for each idx in parallel using GNU Parallel..."
 	@set -e; \
-	tail -n +2 $(CSV_RESULTS) | awk -F, '{print $$1}' | parallel -j 50% python3 src/python/replay_merge.py --merges_csv $(CSV_RESULTS) -delete_workdir --idx {}
+	tail -n +2 $(CSV_RESULTS) | awk -F, '{print $$1}' | parallel --halt now,fail=1 -j 50% python3 src/python/replay_merge.py --merges_csv $(CSV_RESULTS) -delete_workdir --idx {}
 
 protect-repos:
 	find repos -mindepth 1 -type d -exec chmod a-w {} +

README.md

@@ -215,6 +215,5 @@ To run style checking run `make style`.
 ## Comparing merge algorithms
 
 To investigate differences between two mergers:
- * edit file `src/python/select_from_results.py` to reflect the differences you are interested in.
- * run `src/python/select_from_results.py` to create a .csv database containing only the differences.
- * run `src/python/replay_merge.py --merges_csv CSV_FILE --idx INDEX` (maybe add `-test`) for the index of the merge you are interested in.
+ * run `src/python/select_from_results.py` to produce a CSV file containing only the differences you are interested in.  See its [documentation (at top of file)](src/python/select_from_results.py) for how to run it.
+ * run `src/python/replay_merge.py --idx INDEX` (maybe add `-test`) for the index of the merge you are interested in.

run.sh

@@ -23,6 +23,13 @@ OUT_DIR="results/$RUN_NAME"
 N_MERGES=$3
 CACHE_DIR="${4}"
 
+# Do not run this script on MacOS.
+backend=$(uname -s)
+if [ "$backend" = "Darwin" ]; then
+    echo "Error: MacOS is not supported. Please run the script on a Linux machine. This is due to the use of readarray in certain merge tools."
+    exit 1
+fi
+
 comparator_flags=""
 no_timing=false
 only_plotting=false
@@ -52,7 +59,15 @@ while [ $# -gt 0 ]; do
   shift
 done
 
-PATH=$(pwd)/src/scripts/merge_tools/:$PATH
+# Check if src/scripts/merge_tools/merging is present
+if [ ! -d src/scripts/merge_tools/merging ]; then
+    echo "Error: src/scripts/merge_tools/merging is missing. This is a submodule that is required for the script to run."
+    echo "Please run 'git submodule update --init' to fetch the submodule."
+    exit 1
+fi
+
+PATH=$(pwd)/src/scripts/merge_tools:$PATH
+PATH=$(pwd)/src/scripts/merge_tools/merging/src/main/sh:$PATH
 export PATH
 
 GIT_CONFIG_GLOBAL=$(pwd)/.gitconfig
@@ -72,6 +87,13 @@ if [ -z "${JAVA17_HOME:+isset}" ] ; then echo "JAVA17_HOME is not set"; exit 1;
 if [ -z "${machine_id:+isset}" ] ; then machine_id=0; fi
 if [ -z "${num_machines:+isset}" ] ; then num_machines=1; fi
 
+export JAVA_HOME=$JAVA17_HOME
+if [ ! -f ./src/scripts/merge_tools/merging/.git ] ; then
+    git submodule update --init --recursive
+fi
+
+(cd ./src/scripts/merge_tools/merging && ./gradlew shadowJar)
+
 echo "Machine ID: $machine_id"
 echo "Number of machines: $num_machines"
 echo "Output directory: $OUT_DIR"
@@ -93,6 +115,7 @@ if [ "$only_plotting" = true ]; then
     exit 0
 fi
 
+export JAVA_HOME=$JAVA11_HOME
 ./gradlew -q assemble -g ../.gradle/
 
 mkdir -p "$OUT_DIR"
@@ -126,13 +149,25 @@ java -cp build/libs/astmergeevaluation-all.jar \
     "$OUT_DIR/merges"
 
 # Sample 5*<n_merges> merges
-read -ra merge_comparator_flags <<<"${comparator_flags}"
-python3 src/python/merges_sampler.py \
-    --repos_head_passes_csv "$OUT_DIR/repos_head_passes.csv" \
-    --merges_path "$OUT_DIR/merges/" \
-    --output_dir "$OUT_DIR/merges_sampled/" \
-    --n_merges "$((5 * "$N_MERGES"))" \
-    "${merge_comparator_flags[@]}"
+total_merges=$((5 * N_MERGES))
+
+# Ensure comparator_flags is set, but default to an empty array if not
+if [[ -n "${comparator_flags}" ]]; then
+    read -ra merge_comparator_flags <<< "${comparator_flags}"
+    python3 src/python/merges_sampler.py \
+        --repos_head_passes_csv "$OUT_DIR/repos_head_passes.csv" \
+        --merges_path "$OUT_DIR/merges/" \
+        --output_dir "$OUT_DIR/merges_sampled/" \
+        --n_merges "$total_merges" \
+        "${merge_comparator_flags[@]}"
+else
+    echo "Warning: 'comparator_flags' is empty, continuing without additional flags."
+    python3 src/python/merges_sampler.py \
+        --repos_head_passes_csv "$OUT_DIR/repos_head_passes.csv" \
+        --merges_path "$OUT_DIR/merges/" \
+        --output_dir "$OUT_DIR/merges_sampled/" \
+        --n_merges "$total_merges"
+fi
 
 python3 src/python/split_repos.py \
     --repos_csv "$OUT_DIR/repos_head_passes.csv" \
@@ -165,6 +200,17 @@ if [ "$no_timing" = false ]; then
         --n_timings 3 \
         --cache_dir "$CACHE_DIR"
     extra_args+=(--timed_merges_path "$OUT_DIR/merges_timed/")
+
+    python3 src/python/latex_output.py \
+        --run_name "$RUN_NAME" \
+        --merges_path "$OUT_DIR/merges/" \
+        --tested_merges_path "$OUT_DIR/merges_tested/" \
+        --analyzed_merges_path "$OUT_DIR/merges_analyzed/" \
+        --timed_merges_path "$OUT_DIR/merges_timed/" \
+        --full_repos_csv "$REPOS_CSV_WITH_HASHES" \
+        --repos_head_passes_csv "$OUT_DIR/repos_head_passes.csv" \
+        --n_merges "$N_MERGES" \
+        --output_dir "$OUT_DIR"
 fi
 
 python3 src/python/latex_output.py \

src/python/replay_merge.py

@@ -1,6 +1,11 @@
 #!/usr/bin/env python3
 # -*- coding: utf-8 -*-
-"""Replay merges and their test results"""
+"""Replay merges and their test results.
+
+Typical usage:
+  replay_merge.py --idx INDEX
+where INDEX is, for example, 38-192 .
+"""
 import argparse
 import os
 import sys
@@ -50,7 +55,7 @@ def delete_workdirs(results_df: pd.DataFrame) -> None:
     logger.info("Workdirs deleted")
 
 
-# pylint: disable=too-many-arguments, too-many-locals
+# pylint: disable=too-many-arguments, too-many-locals, too-many-branches, too-many-statements
 def merge_replay(
     merge_idx: str,
     repo_slug: str,
@@ -95,9 +100,12 @@ def merge_replay(
                 logger.info(
                     f"workdir {WORKDIR_DIRECTORY / workdir} already exists for idx: {merge_idx}"
                 )
-                answer = input(
-                    f"workdir {workdir} exists for idx: {merge_idx}. Delete it? (y/n)"
-                )
+                if delete_workdir:
+                    answer = "y"
+                else:
+                    answer = input(
+                        f"workdir {workdir} exists for idx: {merge_idx}. Delete it? (y/n)"
+                    )
                 if answer == "y":
                     shutil.rmtree(WORKDIR_DIRECTORY / workdir)
                 else:
@@ -276,6 +284,11 @@ def merge_replay(
         help="Don't check the fingerprint of a merge",
         action="store_true",
     )
+    parser.add_argument(
+        "-skip_build",
+        help="Build the merge tool",
+        action="store_false",
+    )
     parser.add_argument(
         "-create_artifacts",
         help="Create artifacts",
@@ -297,6 +310,19 @@ def merge_replay(
         logger.info("Testing the replay of a merge")
     if args.create_artifacts:
         logger.info("Creating artifacts after replaying the merges")
+    if args.skip_build:
+        logger.info("Building merge tool")
+
+    os.environ["PATH"] += os.pathsep + os.path.join(
+        os.getcwd(), "src/scripts/merge_tools/merging/src/main/sh/"
+    )
+    os.environ["PATH"] += os.pathsep + os.path.join(
+        os.getcwd(), "src/scripts/merge_tools"
+    )
+    os.environ["GIT_CONFIG_GLOBAL"] = os.getcwd() + "/.gitconfig"
+    if not args.skip_build:
+        os.system("cd src/scripts/merge_tools/merging && ./gradlew -q shadowJar")
+    os.system("git submodule update --init")
 
     df = pd.read_csv(args.merges_csv, index_col="idx")
 

src/python/select_from_results.py

@@ -13,9 +13,14 @@
 Here are example invocations:
   select_from_results.py '(gitmerge_ort == "Merge_failed") and (spork != "Merge_failed")'
   select_from_results.py '(gitmerge_ort == "Merge_failed") != (spork == "Merge_failed")'
+
+The resulting .csv is useful for manual examination but cannot be passed to
+`replay_merge.py` because that requires a .csv file with all tools and all
+fingerprints.
 """
 
 import argparse
+import os
 from os import system
 import re
 import tempfile
@@ -25,9 +30,9 @@
 def columns_in_query(query):
     """Returns all the identifiers used in the query."""
     result = re.findall(r"""(?<!['"])\b[A-Za-z][A-Za-z_]*\b(?!['"])""", query)
-    if "and" in result:
+    while "and" in result:
         result.remove("and")
-    if "or" in result:
+    while "or" in result:
         result.remove("or")
     return result
 
@@ -44,8 +49,11 @@ def main():
         description="Outputs a subset of the results, to standard out",
     )
     parser.add_argument("query")
+    scriptdir = os.path.dirname(os.path.realpath(__file__))
     parser.add_argument(
-        "--input", action="store", default="../../results/combined/result.csv"
+        "--input",
+        action="store",
+        default=scriptdir + "/" + "../../results/combined/result.csv",
     )
     parser.add_argument("columns", nargs=argparse.REMAINDER)
     args = parser.parse_args()
@@ -57,16 +65,26 @@ def main():
 
     # Select some columns
     columns_to_select = (
-        ["idx", "repo-idx", "merge-idx", "branch_name", "merge", "left", "right"]
+        [
+            "idx",
+            "repo-idx",
+            "merge-idx",
+            "branch_name",
+            "merge",
+            "left",
+            "left_tree_fingerprint",
+            "right",
+            "right_tree_fingerprint",
+        ]
         + columns_in_query(args.query)
         + args.columns
+        + ["repository"]
     )
     df = df[columns_to_select]
 
     # Gross way to produce output to standard out
-    with tempfile.TemporaryFile() as tmpfile:
+    with tempfile.NamedTemporaryFile() as tmpfile:
         df.to_csv(tmpfile)
-        print(tmpfile.name)
         system("cat " + tmpfile.name)
 
 

src/scripts/merge_tools/gitmerge_ort_imports.sh

@@ -6,15 +6,25 @@ MERGE_SCRIPTS_DIR="$(cd "$(dirname "$0")" && pwd -P)"
 clone_dir=$1
 branch1=$2
 branch2=$3
-strategy="-s ort"
-if "$MERGE_SCRIPTS_DIR"/gitmerge.sh "$clone_dir" "$branch1" "$branch2" "$strategy"; then
-  exit 0
-fi
+
+# shellcheck disable=SC2153 # "JAVA17_HOME" is not a misspelling of "JAVA_HOME"
+export JAVA_HOME="$JAVA17_HOME"
 
 cd "$clone_dir" || exit 1
-if ! "$MERGE_SCRIPTS_DIR"/resolve-import-conflicts; then
-  echo "Conflict"
-  exit 1
+
+git checkout "$branch1" --force
+
+attributes_file=".git/info/attributes"
+echo "*.java merge=merge-java" >> "$attributes_file"
+git config --local merge.merge-java.name "Merge Java files"
+git config --local merge.merge-java.driver 'java-merge-driver.sh "%A" "%O" "%B"'
+
+git merge --no-edit "$branch2"
+retVal=$?
+
+# report conflicts
+if [ "$retVal" -ne 0 ]; then
+    echo "Conflict"
 fi
 
-exit 0
+exit "$retVal"

src/scripts/merge_tools/gitmerge_ort_imports_ignorespace.sh

@@ -6,15 +6,26 @@ MERGE_SCRIPTS_DIR="$(cd "$(dirname "$0")" && pwd -P)"
 clone_dir=$1
 branch1=$2
 branch2=$3
-strategy="-s ort -Xignore-space-change"
-if "$MERGE_SCRIPTS_DIR"/gitmerge.sh "$clone_dir" "$branch1" "$branch2" "$strategy"; then
-  exit 0
-fi
+strategy="-Xignore-space-change"
+
+# shellcheck disable=SC2153 # "JAVA17_HOME" is not a misspelling of "JAVA_HOME"
+export JAVA_HOME="$JAVA17_HOME"
 
 cd "$clone_dir" || exit 1
-if ! "$MERGE_SCRIPTS_DIR"/resolve-import-conflicts; then
-  echo "Conflict"
-  exit 1
+
+git checkout "$branch1" --force
+
+attributes_file=".git/info/attributes"
+echo "*.java merge=merge-java" >> "$attributes_file"
+git config --local merge.merge-java.name "Merge Java files"
+git config --local merge.merge-java.driver 'java-merge-driver.sh "%A" "%O" "%B"'
+
+git merge --no-edit "$strategy" "$branch2"
+retVal=$?
+
+# report conflicts
+if [ "$retVal" -ne 0 ]; then
+    echo "Conflict"
 fi
 
-exit 0
+exit "$retVal"