TST: Use xlsx format instead of xls.

The xls conversion was corrupting some equations, so switching to
use Excel produced xlsx.

tests/risk/answer_key.py

@@ -34,12 +34,10 @@ DIR = os.path.dirname(os.path.realpath(__file__))
 ANSWER_KEY_CHECKSUMS_PATH = os.path.join(DIR, 'risk-answer-key-checksums')
 ANSWER_KEY_CHECKSUMS = open(ANSWER_KEY_CHECKSUMS_PATH, 'r').read().splitlines()
 
-ANSWER_KEY_PATH = os.path.join(DIR, 'risk-answer-key.xls')
-
-ANSWER_KEY_EXISTS = os.path.exists(ANSWER_KEY_PATH)
+ANSWER_KEY_PATH = os.path.join(DIR, 'risk-answer-key.xlsx')
 
 ANSWER_KEY_DL_TEMPLATE = """
-https://s3.amazonaws.com/zipline-test-data/risk/{md5}+/risk-answer-key.xls
+https://s3.amazonaws.com/zipline-test-data/risk/{md5}/risk-answer-key.xlsx
 """.strip()
 
 
@@ -49,29 +47,23 @@ def ensure_latest_answer_key():
 
     Logic for determining what and when to download is as such:
 
-    - If there is no local xls file, then get the lastest answer key,
+    - If there is no local spreadsheet file, then get the lastest answer key,
     as defined by the last row in the checksum file.
-    - If there is a local xls file:
-    -- If the xls's checksum is in the checksum file:
-    --- If the xls's checksum does not match the latest, then grab the
+    - If there is a local spreadsheet file:
+    -- If the spreadsheet's checksum is in the checksum file:
+    --- If the spreadsheet's checksum does not match the latest, then grab the
     the latest checksum and replace the local checksum file.
-    --- If the xls's checksum matches the latest, then skip download, and
-    use the local xls as a cached copy.
-    -- If the xls's checksum is not in the checksum file, then leave the
-    local file alone, assuming that the local xls's md5 is not in the list due
-    to local modifications during development.
+    --- If the spreadsheet's checksum matches the latest, then skip download,
+    and use the local spreadsheet as a cached copy.
+    -- If the spreadsheet's checksum is not in the checksum file, then leave
+    the local file alone, assuming that the local xls's md5 is not in the list
+    due to local modifications during development.
 
     It is possible that md5's could collide, if that is ever case, we should
     then find an alternative naming scheme.
 
-    The xls answer sheet is not kept in SCM, because its size is on the order
-    of 20MB, and every edit would increase the repo size.
-
-    xlsx and ods have smaller outputs and could be more friendly to SCM, but:
-    - not using xlsx, because currently the xlsx that is generated by
-    LibreOffice is not readable by the xldr module.
-    - not using ods, because of the lack of a module as facile as xldr for
-    extracting the data from the ods format.
+    The spreadsheet answer sheet is not kept in SCM, as every edit would
+    increase the repo size by the file size, since it is treated as a binary.
     """
 
     answer_key_dl_checksum = None

tests/risk/risk-answer-key-checksums

@@ -1 +1,2 @@
 3ac0773c4be4e9e5bacd9c6fa0e03e15
+3a5fae958c8bac684f1773fa8dff7810