[xcl-lens] Update verbose mode

projects/xcl-lens/src/xcl_lens/parser/rccl.py

@@ -159,7 +159,7 @@ class RcclLogParser:
             df_ib.sort_values(by=["host", "rank", "hca_no", "hca_id"], inplace=True)
             df_ib = df_ib[["host", "rank", "hca_no", "hca_id", "gdr"]]
             if not self._verbose:
-                df_ib = df_ib.drop(columns=["host"])
+                df_ib = df_ib.drop(columns=["host", "rank"])
                 df_ib.drop_duplicates(inplace=True)
             print(df_ib.to_string(index=False))
             print()
@@ -205,14 +205,14 @@ class RcclLogParser:
             )
             df_gpu = df_gpu[["host", "rank", "gpu", "hca_no", "distance", "max_distance", "r/w"]]
             if not self._verbose:
-                df_gpu = df_gpu.drop(columns=["host"])
+                df_gpu = df_gpu.drop(columns=["host", "rank"])
                 df_gpu.drop_duplicates(inplace=True)
             print(df_gpu.to_string(index=False))
             print()
         else:
             print("  (No data found)\n")
 
-    def _extract_and_print(self, title, filter_func, fields, mandatory, sort_cols, move_rank=True):
+    def _extract_and_print(self, title, filter_func, fields, mandatory, verbose_cols, sort_cols, move_rank=True):
         """
         Generic function to extract structured data from log lines and print as a table.
 
@@ -235,6 +235,7 @@ class RcclLogParser:
                    - col_name: Name of the DataFrame column
                    - value_pattern: Regex pattern to validate/extract the field value
             mandatory: List of column names that must not be NaN (drop rows missing these)
+            verbose_cols: List of column names to keep when not verbose
             sort_cols: List of column names to sort by (in order)
             move_rank: If True, move "rank" column to front and "protocol" to second if present
         """
@@ -254,6 +255,10 @@ class RcclLogParser:
                 rf"\b{pattern}\s+({val_pattern})", expand=False
             )
 
+        # Drop verbose columns if not verbose
+        if not self._verbose:
+            df = df.drop(columns=verbose_cols, errors="ignore")
+
         # Convert numeric fields to appropriate types
         numeric_columns = [
             "Pattern",
@@ -282,9 +287,6 @@ class RcclLogParser:
         mandatory = [c for c in mandatory if c in df.columns]
         df.dropna(subset=mandatory, inplace=True)  # Remove rows missing mandatory fields
         df.drop(columns=["raw_log"], inplace=True)  # No longer need raw log
-
-        if not self._verbose and "host" in df.columns:
-            df = df.drop(columns=["host"])
         df.drop_duplicates(inplace=True)  # Deduplicate identical records
 
         if df.empty:
@@ -293,19 +295,10 @@ class RcclLogParser:
 
         # Reorder columns for better readability
         if move_rank:
-            cols = df.columns.tolist()
-            if "host" in cols:
-                cols.remove("host")
-            cols.remove("rank")
-            # Move protocol to second position if present
-            if "protocol" in cols:
-                cols.remove("protocol")
-                cols.insert(0, "protocol")
-            # Always move host and rank to front
-            cols.insert(0, "rank")
-            if "host" in cols:
-                cols.insert(0, "host")
-            df = df[cols]
+            target_order = ["host", "rank", "protocol"]
+            leading_cols = [c for c in target_order if c in df.columns]
+            remaining_cols = [c for c in df.columns if c not in leading_cols]
+            df = df[leading_cols + remaining_cols]
 
         # Sort the data
         sort_cols = [c for c in sort_cols if c in df.columns]
@@ -329,11 +322,38 @@ class RcclLogParser:
             filter_func=lambda c: "Pattern" in c and "crossNic" in c,
             fields=self.graph_info_fields,
             mandatory=["Pattern"],
-            sort_cols=["rank", "Pattern"],
+            verbose_cols=["host", "rank"],
+            sort_cols=["host", "rank", "Pattern"],
+        )
+
+    def _report_collective_transfers(self):
+        self._extract_and_print(
+            title="Unique Ring/Tree Transfers",
+            filter_func=lambda c: "protocol" in c and "nbytes" in c,
+            fields=self.cl_transfer_fields,
+            mandatory=["protocol", "nbytes"],
+            verbose_cols=["host", "rank"],
+            sort_cols=["host", "rank", "nbytes", "protocol", "nchannels"],
+        )
+
+    def _report_p2p_transfers(self):
+        self._extract_and_print(
+            title="Unique P2P Transfers",
+            filter_func=lambda c: "p2p :" in c and "send rank" in c,
+            fields=self.p2p_fields,
+            mandatory=["local", "send", "recv"],
+            verbose_cols=["host", "rank", "local", "send", "recv"],
+            sort_cols=["host", "rank", "protocol", "local", "send", "recv"],
         )
 
     def _report_channel_transport_info(self):
         print("===> Channel Transport Info:\n")
+
+        if not self._verbose:
+            print("  (Skipped because verbose mode is not enabled)")
+            print()
+            return
+
         data = []
 
         # Match pattern: Channel 00/0 : 2[5d000] -> 1[56000] [send] via NET/IB/6/GDRDMA
@@ -369,26 +389,6 @@ class RcclLogParser:
 
         df = pd.DataFrame(data)
         df.sort_values(by=["host", "rank", "channel", "sender", "receiver"], inplace=True)
-        if not self._verbose:
-            df = df.drop(columns=["host", "channel", "sender", "receiver"])
         df.drop_duplicates(inplace=True)
         print(df.to_string(index=False))
         print()
-
-    def _report_collective_transfers(self):
-        self._extract_and_print(
-            title="Unique Ring/Tree Transfers",
-            filter_func=lambda c: "protocol" in c and "nbytes" in c,
-            fields=self.cl_transfer_fields,
-            mandatory=["protocol", "nbytes"],
-            sort_cols=["rank", "nbytes", "protocol", "nchannels"],
-        )
-
-    def _report_p2p_transfers(self):
-        self._extract_and_print(
-            title="Unique P2P Transfers",
-            filter_func=lambda c: "p2p :" in c and "send rank" in c,
-            fields=self.p2p_fields,
-            mandatory=["local", "send", "recv"],
-            sort_cols=["rank", "protocol", "local", "send", "recv"],
-        )