add important considerations at the top of methology

justfile

@@ -4,52 +4,8 @@
 set dotenv-load
 
 # variables
-## n2
-#instance_name := "ibis-bench-2"
-#instance_type := "c3-standard-4"
-# 
-#instance_name := "ibis-bench-3"
-#instance_type := "c3-standard-8"
-# 
-#instance_name := "ibis-bench-4"
-#instance_type := "c3-standard-22"
-# 
-#instance_name := "ibis-bench-5"
-#instance_type := "c3-standard-44"
-# 
-# # n2d
-instance_name := "ibis-bench-6"
+instance_name := "ibis-bench"
 instance_type := "n2d-standard-2"
-# 
-#instance_name := "ibis-bench-7"
-#instance_type := "n2d-standard-4"
-# 
-#instance_name := "ibis-bench-8"
-#instance_type := "n2d-standard-8"
-# 
-#instance_name := "ibis-bench-9"
-#instance_type := "n2d-standard-16"
-# 
-#instance_name := "ibis-bench-10"
-#instance_type := "n2d-standard-32"
-# 
-# # n4
-#instance_name := "ibis-bench-11"
-#instance_type := "n2-standard-2"
-# 
-#instance_name := "ibis-bench-12"
-#instance_type := "n2-standard-4"
-# 
-#instance_name := "ibis-bench-13"
-#instance_type := "n2-standard-8"
-# 
-#instance_name := "ibis-bench-14"
-#instance_type := "n2-standard-16"
-# 
-#instance_name := "ibis-bench-15"
-#instance_type := "n2-standard-32"
-
-### --- ####
 instance_zone := "us-central1-b"
 
 gen_scale_factors := "-s 1 -s 8 -s 16 -s 32 -s 64 -s 128" 

pages/methodology.py

@@ -16,6 +16,20 @@
 """
 st.markdown(methodology)
 
+methodology = """
+## Important considerations
+
+Noted in more detail below (with code as the source of truth), but some important considerations include:
+
+- TPC-H data is generated via DuckDB as Parquet files, with non-integers as decimals
+- data is read in with decimals converted to floats for each query, by Ibis (via the corresponding backend) and Polars
+    - additionally, hive-style partitioned columns are dropped
+- queries are run 3 times and the average is taken
+- all data is available in a public `gs://ibis-bench/` bucket for analysis (you can take a minimum instead of an average, for example)
+- some Polars TPC-H queries were recently re-written; we will update `ibis-bench` and re-run soon including these and using newer versions of libraries
+"""
+st.markdown(methodology)
+
 methodology = """
 ## Data generation