yuclid 0.1.2__py3-none-any.whl → 0.1.4__py3-none-any.whl

yuclid/__init__.py

@@ -1 +1 @@
-__version__ = "0.1.0"
+__version__ = "0.1.4"

yuclid/cli.py

@@ -109,13 +109,6 @@ def get_parser():
         nargs="*",
         help="Normalize all values w.r.t. a single reference",
     )
-    # plot_parser.add_argument(
-    #     "-S",
-    #     "--norm-scope",
-    #     choices=["global", "local"],
-    #     default="local",
-    #     help="Normalization scope",
-    # )
     plot_parser.add_argument(
         "-r",
         "--norm-reverse",

yuclid/plot.py

@@ -1,6 +1,4 @@
 from yuclid.log import report, LogLevel
-import yuclid.cli
-import matplotlib.gridspec as gridspec
 import matplotlib.lines as mlines
 import matplotlib.pyplot as plt
 import yuclid.spread as spread
@@ -8,14 +6,10 @@ import seaborn as sns
 import pandas as pd
 import numpy as np
 import scipy.stats
-import subprocess
-import threading
+import yuclid.cli
 import itertools
 import pathlib
-import hashlib
-import time
 import math
-import sys
 
 
 def get_current_config(ctx):
@@ -168,8 +162,6 @@ def generate_dataframe(ctx):
 
     if len(dfs) == 0:
         report(LogLevel.ERROR, "no valid source of data")
-        ctx["alive"] = False
-        sys.exit(1)
 
     df = pd.concat(dfs)
 
@@ -208,7 +200,6 @@ def rescale(ctx):
     args = ctx["args"]
     for y in args.y:
         df[y] = df[y] * args.rescale
-    ctx["df"] = df
 
 
 def draw(fig, ax, cli_args):
@@ -254,34 +245,6 @@ def generate_space(ctx):
     )
 
 
-def file_monitor(ctx):
-    current_hash = None
-    last_hash = None
-    while ctx["alive"]:
-        try:
-            current_hash = ""
-            for file in ctx["local_files"]:
-                with open(file, "rb") as f:
-                    current_hash += hashlib.md5(f.read()).hexdigest()
-        except FileNotFoundError:
-            current_hash = None
-        if current_hash != last_hash:
-            generate_dataframe(ctx)
-            rescale(ctx)
-            generate_space(ctx)
-            compute_ylimits(ctx)
-            space_columns = ctx["df"].columns.difference([ctx["y_axis"]])
-            sizes = ["{}={}".format(d, ctx["df"][d].nunique()) for d in space_columns]
-            missing = compute_missing(ctx)
-            report(LogLevel.INFO, "space sizes", " | ".join(sizes))
-            if len(missing) > 0:
-                report(LogLevel.WARNING, f"at least {len(missing)} missing experiments")
-            update_table(ctx)
-            update_plot(ctx)
-        last_hash = current_hash
-        time.sleep(1)
-
-
 def update_table(ctx):
     ax_table = ctx["ax_table"]
     free_dims = ctx["free_dims"]
@@ -317,32 +280,6 @@ def is_remote(file):
     return "@" in file
 
 
-def sync_files(ctx):
-    args = ctx["args"]
-    valid_files = ctx["valid_files"]
-    jobs = []
-    for file in valid_files:
-        if is_remote(file):
-            mirror = get_local_mirror(file)
-            proc = subprocess.run(["scp", file, mirror])
-            if proc.returncode != 0:
-                report(LogLevel.ERROR, f"scp transfer failed for {file}")
-                sys.exit(1)
-            jobs.append((file, mirror))
-
-    def rsync(src, dst):
-        while ctx["alive"]:
-            subprocess.run(
-                ["rsync", "-z", "--checksum", src, dst],
-                stdout=subprocess.DEVNULL,
-                stderr=subprocess.DEVNULL,
-            )
-            time.sleep(args.rsync_interval)
-
-    for job in jobs:
-        threading.Thread(target=rsync, daemon=True, args=job).start()
-
-
 def fontsize_to_y_units(ctx, fontsize):
     fig = ctx["fig"]
     ax = ctx["ax_plot"]
@@ -952,13 +889,9 @@ def validate_args(ctx):
 
 
 def start_gui(ctx):
-    ctx["alive"] = True
-
     update_plot(ctx)
     update_table(ctx)
-    threading.Thread(target=file_monitor, daemon=True, args=(ctx,)).start()
     report(LogLevel.INFO, "application running")
-    time.sleep(1.0)  # wait for the GUI to initialize
     plt.show()
 
 
@@ -995,12 +928,48 @@ def compute_ylimits(ctx):
     ctx["top"] = top
 
 
+def generate_derived_metrics(ctx):
+    args = ctx["args"]
+    df = ctx["df"]
+
+    # derived metrics are any -y value with a ":"
+    derived_metrics = dict()
+    new_ys = []
+    for y in args.y:
+        if ":" in y:
+            name, func = y.split(":")
+            derived_metrics[name.strip()] = func.strip()
+            new_ys.append(name.strip())
+        else:
+            new_ys.append(y)
+
+    for name, func in derived_metrics.items():
+        try:
+            # replace column names in the expression with df[column_name] syntax
+            expression = func
+            for col in df.columns:
+                if col in expression:
+                    expression = expression.replace(col, f"df['{col}']")
+
+            df[name] = eval(expression)
+        except Exception as e:
+            hint = "maybe you misspelled a column name"
+            report(
+                LogLevel.ERROR,
+                f"failed to evaluate derived metric '{name}'",
+                hint=hint,
+            )
+            continue
+
+    args.y = new_ys
+
+
 def launch(args):
-    ctx = {"args": args, "alive": True}
+    ctx = {"args": args}
     validate_files(ctx)
     locate_files(ctx)
-    sync_files(ctx)
     generate_dataframe(ctx)
+    generate_derived_metrics(ctx)
     validate_args(ctx)
     rescale(ctx)
     generate_space(ctx)

yuclid/run.py

@@ -110,29 +110,36 @@ def load_json(f):
 
 
 def aggregate_input_data(settings):
-    data = None
+    data = {
+        "env": {},
+        "setup": {"global": [], "point": []},
+        "space": {},
+        "trials": [],
+        "metrics": [],
+        "presets": {},
+        "order": [],
+    }
 
     for file in settings["inputs"]:
         with open(file, "r") as f:
             current = normalize_data(load_json(f))
-            if data is None:
-                data = current
-                continue
             for key, val in current.items():
-                if isinstance(data[key], list):
+                if key in ["env", "space", "presets"]:
+                    data[key].update(val)
+                elif key in ["trials", "metrics", "order"]:
                     data[key].extend(val)
-                elif isinstance(data[key], dict):
-                    if key == "space":
-                        for subkey, subval in val.items():
-                            if data[key].get(subkey) is None:
-                                data[key][subkey] = subval
-                            else:
-                                data[key].setdefault(subkey, []).extend(subval)
-                    else:
-                        data[key].update(val)
-
-            order = data.get("order", []) + current.get("order", [])
-            data["order"] = remove_duplicates(order)
+                elif key == "setup":
+                    for subkey, subval in val.items():
+                        if data[key].get(subkey) is None:
+                            # undefined dimensions are overridden
+                            data[key][subkey] = subval
+                        else:
+                            data[key].setdefault(subkey, []).extend(subval)
+                elif key == "setup":
+                    data[key]["setup"]["global"] += val["setup"]["global"]
+                    data[key]["setup"]["point"] += val["setup"]["point"]
+
+    data["order"] = remove_duplicates(data["order"])
 
     if len(data["trials"]) == 0:
         report(LogLevel.FATAL, "no valid trials found")
@@ -735,10 +742,12 @@ def run_point_trials(settings, data, execution, f, i, point):
         )
 
     i_padded = str(i).zfill(len(str(execution["subspace_size"])))
-    
+
     for j, trial in enumerate(compatible_trials):
         point_id = os.path.join(
-            settings["temp_dir"], settings["now"], f"{i_padded}." + point_to_string(point) + f"_trial{j}"
+            settings["temp_dir"],
+            settings["now"],
+            f"{i_padded}." + point_to_string(point) + f"_trial{j}",
         )
 
         command = substitute_global_yvars(trial["command"], execution["subspace"])
@@ -780,7 +789,8 @@ def run_point_trials(settings, data, execution, f, i, point):
             capture_output=True,
             env=execution["env"],
         )
-        if command_output.returncode != 0:
+
+        def complain():
             hint = "check the following files for more details:\n"
             hint += f"{point_id}.out\n{point_id}.err\n{point_id}.tmp"
             report(
@@ -791,6 +801,9 @@ def run_point_trials(settings, data, execution, f, i, point):
                 ),
                 hint=hint,
             )
+
+        if command_output.returncode != 0:
+            complain()
         else:
             output_lines = command_output.stdout.strip().split("\n")
 
@@ -798,7 +811,10 @@ def run_point_trials(settings, data, execution, f, i, point):
                 try:
                     return int(x)
                 except ValueError:
-                    return float(x)
+                    try:
+                        return float(x)
+                    except ValueError:
+                        complain()
 
             collected_metrics[metric["name"]] = [
                 int_or_float(line) for line in output_lines
@@ -871,6 +887,20 @@ def validate_execution(execution, data):
                 point_to_string(point),
                 hint="try relaxing your trial conditions or adding more trials.",
             )
+        if len(execution["metrics"] or []) > 0:
+            compatible_metric_names = {m["name"] for m in compatible_metrics}
+            incompatible = [
+                m for m in execution["metrics"] if m not in compatible_metric_names
+            ]
+            if len(incompatible) > 0:
+                report(
+                    LogLevel.ERROR,
+                    "some metrics are not compatible with {}".format(
+                        point_to_string(point)
+                    ),
+                    ", ".join(incompatible),
+                    hint="try relaxing your metric conditions or adding more metrics.",
+                )
 
 
 def get_compatible_trials_and_metrics(data, point, execution):
@@ -911,25 +941,12 @@ def run_subspace_trials(settings, data, execution):
                 compatible_trials, compatible_metrics = (
                     get_compatible_trials_and_metrics(data, point, execution)
                 )
-                if len(compatible_trials) == 0:
-                    report(
-                        LogLevel.ERROR,
-                        point_to_string(point),
-                        "no compatible trials found",
-                    )
-                elif len(compatible_metrics) == 0:
-                    report(
-                        LogLevel.ERROR,
-                        point_to_string(point),
-                        "no compatible metrics found",
-                    )
-                else:
-                    report(
-                        LogLevel.INFO,
-                        get_progress(i, execution["subspace_size"]),
-                        "dry run",
-                        point_to_string(point),
-                    )
+                report(
+                    LogLevel.INFO,
+                    get_progress(i, execution["subspace_size"]),
+                    "dry run",
+                    point_to_string(point),
+                )
     else:
         output_dir = os.path.dirname(settings["output"])
         if output_dir and not os.path.exists(output_dir):
@@ -1122,7 +1139,7 @@ def normalize_point_setup(point_setup, space):
         report(LogLevel.FATAL, "point setup must be a string or a list")
 
     # check validity of 'on' fields
-    for item in point_setup:
+    for item in normalized_items:
         if not isinstance(item["on"], (list, type(None))):
             report(LogLevel.FATAL, "point setup 'on' must be a list or None")
         for dim in item["on"]:
@@ -1139,7 +1156,7 @@ def normalize_point_setup(point_setup, space):
                 )
 
     # check validity of 'parallel' fields
-    for item in point_setup:
+    for item in normalized_items:
         parallel = item["parallel"]
         if not isinstance(parallel, (bool, list)):
             report(LogLevel.FATAL, "point setup 'parallel' must be a boolean or a list")

{yuclid-0.1.2.dist-info → yuclid-0.1.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: yuclid
-Version: 0.1.2
+Version: 0.1.4
 Summary: Run experiments and interactively plot results across combinations of user-specified dimensions
 Author-email: Federico Sossai <federico.sossai@gmail.com>
 Project-URL: Homepage, https://github.com/fsossai/yuclid

yuclid-0.1.4.dist-info/RECORD

@@ -0,0 +1,11 @@
+yuclid/__init__.py,sha256=JMD28FXYHc_TM03visyUSd3UA9FZAaJMRStnfZoq50Y,21
+yuclid/cli.py,sha256=YZzxJty5wlUhCOEELvEcJeQb_lQ1Qc89RG4_s5IyKWU,6224
+yuclid/log.py,sha256=GR_FVfNroumuonKguAPd6H1rKjxJKRc8tAS2sVNTbzE,1655
+yuclid/plot.py,sha256=R6IXw6hHuXYFx1MjTKLCIqBfdNORStVEoDidAr-jEuE,29697
+yuclid/run.py,sha256=s1BGCmYckO2s5TSoKNCb4llZpUouxyooxtVlbqsQNTs,44641
+yuclid/spread.py,sha256=4Ci3nsu8n_dhG-AK2IWHKRElQ8oaGdw14LrgNu79biM,4938
+yuclid-0.1.4.dist-info/METADATA,sha256=Qm5Sw-K-L1VGSEJVBwE_C6Ubjp6JNa55SiH7snsWdPM,673
+yuclid-0.1.4.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+yuclid-0.1.4.dist-info/entry_points.txt,sha256=2AvTtyt5iBnjr6HnjqH_3PeSoq9UzIbT92qivmEbOYA,43
+yuclid-0.1.4.dist-info/top_level.txt,sha256=cL5mb4h_4etwTsqhPvSnoVBXImIzPFGd3rINV1nEjPo,7
+yuclid-0.1.4.dist-info/RECORD,,

yuclid-0.1.2.dist-info/RECORD

@@ -1,11 +0,0 @@
-yuclid/__init__.py,sha256=Pru0BlFBASFCFo7McHdohtKkUtgMPDwbGfyUZlE2_Vw,21
-yuclid/cli.py,sha256=l5WUY6Q6nwg7WRrAAPf5uaspG9zrEPE9BA9v3eYI_vE,6410
-yuclid/log.py,sha256=GR_FVfNroumuonKguAPd6H1rKjxJKRc8tAS2sVNTbzE,1655
-yuclid/plot.py,sha256=RV_bgkFDpOGxw7ankW7QsnBsyrholBtYKKj9jUtBAyM,30836
-yuclid/run.py,sha256=NVvcmLiQkzypgqpRbmI4lttgKSy05hYJKVBEp3fxnpA,44106
-yuclid/spread.py,sha256=4Ci3nsu8n_dhG-AK2IWHKRElQ8oaGdw14LrgNu79biM,4938
-yuclid-0.1.2.dist-info/METADATA,sha256=lCvP9NyDUhHzKnu4e8jCHVTFyAkgps_hH27EuwCL0gk,673
-yuclid-0.1.2.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-yuclid-0.1.2.dist-info/entry_points.txt,sha256=2AvTtyt5iBnjr6HnjqH_3PeSoq9UzIbT92qivmEbOYA,43
-yuclid-0.1.2.dist-info/top_level.txt,sha256=cL5mb4h_4etwTsqhPvSnoVBXImIzPFGd3rINV1nEjPo,7
-yuclid-0.1.2.dist-info/RECORD,,