lectures/_static/lecture_specific/pandas/wb_download.py fetches the World Bank indicator over plain HTTP and writes the response body to disk without checking whether the request succeeded. Raised by Copilot while reviewing the downstream port at QuantEcon/lecture-python-programming.ml#6; it is upstream code, so it belongs here rather than in a translation edition.
Current lines 6-9:
wb_data_query = "http://api.worldbank.org/v2/en/indicator/gc.dod.totl.gd.zs?downloadformat=excel"
r = requests.get(wb_data_query)
with open('gd.xls', 'wb') as output:
output.write(r.content)Three points, in rough order of how much they matter:
No error check. On any non-200 response the error page body is written to gd.xls, and the failure only surfaces further down as an opaque pd.read_excel parse error rather than as the network problem it actually is. r.raise_for_status() turns that into a clear failure at the point of the fault.
Plain HTTP. api.worldbank.org serves HTTPS, so this is a free upgrade — and it is a file the lectures hold up as example code, which is the argument for getting it right beyond the MITM exposure itself.
No timeout. requests.get without timeout blocks indefinitely if the endpoint stops responding, which is an unpleasant failure mode inside a notebook build.
Suggested:
wb_data_query = "https://api.worldbank.org/v2/en/indicator/gc.dod.totl.gd.zs?downloadformat=excel"
r = requests.get(wb_data_query, timeout=30)
r.raise_for_status()
with open('gd.xls', 'wb') as output:
output.write(r.content)Note this script is a static asset rather than an executed cell, so nothing in CI exercises it — worth a manual check that the HTTPS endpoint returns the same workbook before merging.
lectures/_static/lecture_specific/pandas/wb_download.pyfetches the World Bank indicator over plain HTTP and writes the response body to disk without checking whether the request succeeded. Raised by Copilot while reviewing the downstream port at QuantEcon/lecture-python-programming.ml#6; it is upstream code, so it belongs here rather than in a translation edition.Current lines 6-9:
Three points, in rough order of how much they matter:
No error check. On any non-200 response the error page body is written to
gd.xls, and the failure only surfaces further down as an opaquepd.read_excelparse error rather than as the network problem it actually is.r.raise_for_status()turns that into a clear failure at the point of the fault.Plain HTTP.
api.worldbank.orgserves HTTPS, so this is a free upgrade — and it is a file the lectures hold up as example code, which is the argument for getting it right beyond the MITM exposure itself.No timeout.
requests.getwithouttimeoutblocks indefinitely if the endpoint stops responding, which is an unpleasant failure mode inside a notebook build.Suggested:
Note this script is a static asset rather than an executed cell, so nothing in CI exercises it — worth a manual check that the HTTPS endpoint returns the same workbook before merging.