Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Support
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
W
wendelin
Project overview
Project overview
Details
Activity
Releases
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Issues
0
Issues
0
List
Boards
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Analytics
Analytics
CI / CD
Repository
Value Stream
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
Boxiang Sun
wendelin
Commits
fc103ecf
Commit
fc103ecf
authored
Jul 03, 2019
by
Ivan Tyagov
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
Remove test which covers afterwards transformation which is not part of data-lake functionality.
parent
7f5cf88b
Changes
3
Show whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
0 additions
and
570 deletions
+0
-570
bt5/erp5_wendelin_data_lake_ingestion/TestTemplateItem/portal_components/test.erp5.testDataIngestion.py
...lateItem/portal_components/test.erp5.testDataIngestion.py
+0
-432
bt5/erp5_wendelin_data_lake_ingestion/TestTemplateItem/portal_components/test.erp5.testDataIngestion.xml
...ateItem/portal_components/test.erp5.testDataIngestion.xml
+0
-137
bt5/erp5_wendelin_data_lake_ingestion/bt/template_test_id_list
...rp5_wendelin_data_lake_ingestion/bt/template_test_id_list
+0
-1
No files found.
bt5/erp5_wendelin_data_lake_ingestion/TestTemplateItem/portal_components/test.erp5.testDataIngestion.py
deleted
100644 → 0
View file @
7f5cf88b
from
Products.ERP5Type.tests.SecurityTestCase
import
SecurityTestCase
import
string
import
random
import
time
import
json
import
csv
import
os
from
datetime
import
datetime
,
timedelta
import
numpy
as
np
import
math
import
base64
from
Products.ZSQLCatalog.SQLCatalog
import
Query
,
ComplexQuery
from
Products.ERP5Type.Log
import
log
import
hashlib
class
TestDataIngestion
(
SecurityTestCase
):
REFERENCE_SEPARATOR
=
"/"
PART_1
=
REFERENCE_SEPARATOR
+
"001"
PART_2
=
REFERENCE_SEPARATOR
+
"002"
PART_3
=
REFERENCE_SEPARATOR
+
"003"
EOF
=
REFERENCE_SEPARATOR
+
"EOF"
FIF
=
REFERENCE_SEPARATOR
+
"fif"
TXT
=
REFERENCE_SEPARATOR
+
"txt"
CSV
=
REFERENCE_SEPARATOR
+
"csv"
TSV
=
REFERENCE_SEPARATOR
+
"tsv"
GZ
=
REFERENCE_SEPARATOR
+
"gz"
NII
=
REFERENCE_SEPARATOR
+
"nii"
SIZE_HASH
=
REFERENCE_SEPARATOR
+
"fake-size"
+
REFERENCE_SEPARATOR
+
"fake-hash"
SINGLE_INGESTION_END
=
REFERENCE_SEPARATOR
RANDOM
=
REFERENCE_SEPARATOR
+
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
3
)])
CHUNK_SIZE_TXT
=
50000
CHUNK_SIZE_CSV
=
25
REF_PREFIX
=
"fake-supplier"
+
REFERENCE_SEPARATOR
+
"fake-dataset"
+
REFERENCE_SEPARATOR
REF_SUPPLIER_PREFIX
=
"fake-supplier"
+
REFERENCE_SEPARATOR
INGESTION_SCRIPT
=
'HandleFifEmbulkIngestion'
USER
=
'zope'
PASS
=
'roque5'
INVALID
=
"_invalid"
NEW
=
"_NEW"
FALSE
=
"FALSE"
TRUE
=
"TRUE"
def
getTitle
(
self
):
return
"DataIngestionTest"
def
getBusinessTemplateList
(
self
):
return
(
'erp5_base'
,
'erp5_web'
,
'erp5_ingestion_mysql_innodb_catalog'
,
'erp5_ingestion'
,
'erp5_dms'
,
'erp5_wendelin'
,
'erp5_callables'
,
'erp5_core'
)
def
afterSetUp
(
self
):
self
.
context
=
self
.
portal
.
UnitTest_getContext
()
self
.
assertEqual
(
self
.
REFERENCE_SEPARATOR
,
self
.
portal
.
getIngestionReferenceDictionary
()[
"reference_separator"
])
self
.
assertEqual
(
self
.
INVALID
,
self
.
portal
.
getIngestionReferenceDictionary
()[
"invalid_suffix"
])
self
.
assertEqual
(
self
.
EOF
,
self
.
REFERENCE_SEPARATOR
+
self
.
portal
.
getIngestionReferenceDictionary
()[
"split_end_suffix"
])
self
.
assertEqual
(
self
.
SINGLE_INGESTION_END
,
self
.
REFERENCE_SEPARATOR
)
self
.
assertEqual
(
self
.
PART_1
,
self
.
REFERENCE_SEPARATOR
+
self
.
portal
.
getIngestionReferenceDictionary
()[
"split_first_suffix"
])
def
getRandomReference
(
self
):
random_string
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
10
)])
return
'UNIT-TEST-'
+
random_string
def
getIngestionReference
(
self
,
reference
,
extension
):
return
self
.
REF_PREFIX
+
reference
+
extension
def
sanitizeReference
(
self
,
reference
):
ingestion_reference
=
self
.
REFERENCE_SEPARATOR
.
join
(
reference
.
split
(
self
.
REFERENCE_SEPARATOR
)[
1
:])
data_stream
=
self
.
getDataStream
(
ingestion_reference
)
ingestion_id
=
data_stream
.
getId
()
return
ingestion_id
,
ingestion_reference
def
getFullReference
(
self
,
ingestion_reference
,
size
,
hash_value
):
return
self
.
REF_SUPPLIER_PREFIX
+
ingestion_reference
+
self
.
REFERENCE_SEPARATOR
+
self
.
REFERENCE_SEPARATOR
+
str
(
""
)
+
self
.
REFERENCE_SEPARATOR
+
""
def
chunks
(
self
,
l
,
n
):
for
i
in
xrange
(
0
,
len
(
l
),
n
):
yield
l
[
i
:
i
+
n
]
def
generateRawDataBytesAndArray
(
self
):
url
=
'data_stream_module/mne_sample_for_test'
sample_data_stream
=
self
.
context
.
restrictedTraverse
(
url
)
raw_data
,
array
,
json_data
=
self
.
context
.
generateRawData
(
sample_data_stream
)
return
raw_data
,
array
,
json_data
def
getIngestionPolicy
(
self
,
reference
,
ingestion_script
):
ingestion_policy
=
self
.
portal
.
portal_catalog
.
getResultValue
(
portal_type
=
'Ingestion Policy'
,
reference
=
reference
)
if
ingestion_policy
!=
None
:
return
ingestion_policy
ingestion_policy
=
self
.
portal
.
portal_ingestion_policies
.
newContent
(
\
id
=
reference
,
portal_type
=
'Ingestion Policy'
,
reference
=
reference
,
version
=
'001'
,
script_id
=
ingestion_script
)
ingestion_policy
.
validate
()
self
.
tic
()
return
ingestion_policy
def
ingestRequest
(
self
,
requestMethod
,
authentication
,
reference
,
eof
,
data_chunk
,
ingestion
):
encoded_data_chunk
=
base64
.
b64encode
(
data_chunk
)
request
=
self
.
portal
.
REQUEST
request
.
method
=
requestMethod
request
.
auth
=
authentication
request
.
set
(
'reference'
,
reference
+
eof
+
self
.
SIZE_HASH
)
request
.
set
(
'data_chunk'
,
encoded_data_chunk
)
ingestion
.
ingest
()
self
.
tic
()
return
def
getDataIngestion
(
self
,
reference
):
data_ingestion
=
self
.
portal
.
portal_catalog
.
getResultValue
(
portal_type
=
'Data Ingestion'
,
reference
=
reference
)
return
data_ingestion
def
getDataStream
(
self
,
reference
):
data_stream
=
self
.
portal
.
portal_catalog
.
getResultValue
(
portal_type
=
'Data Stream'
,
reference
=
reference
)
return
data_stream
def
getDataAnalysis
(
self
,
reference
):
data_analysis
=
self
.
portal
.
portal_catalog
.
getResultValue
(
portal_type
=
'Data Analysis'
,
reference
=
reference
)
return
data_analysis
def
getDataArray
(
self
,
reference
):
data_array
=
self
.
portal
.
portal_catalog
.
getResultValue
(
portal_type
=
'Data Array'
,
reference
=
reference
)
return
data_array
def
getDataDescriptor
(
self
,
reference
):
data_ingestion
=
None
query
=
ComplexQuery
(
Query
(
simulation_state
=
'stopped'
),
Query
(
simulation_state
=
'delivered'
),
logical_operator
=
"OR"
)
ing_dict
=
{
"query"
:
query
,
"portal_type"
:
"Data Ingestion"
,
"reference"
:
reference
}
ingestions
=
self
.
portal
.
portal_catalog
(
**
ing_dict
)
if
len
(
ingestions
)
==
1
:
data_ingestion
=
ingestions
[
0
]
if
data_ingestion
==
None
:
return
None
url
=
'data_descriptor_module/'
+
data_ingestion
.
getId
()
data_descriptor
=
self
.
context
.
restrictedTraverse
(
url
)
return
data_descriptor
def
manuallyStopIngestionWorkaround
(
self
,
reference
,
now_time
):
try
:
data_ingestion
=
self
.
portal
.
portal_catalog
.
getResultValue
(
portal_type
=
'Data Ingestion'
,
reference
=
reference
)
if
data_ingestion
.
getSimulationState
()
==
"started"
:
data_ingestion
.
stop
()
self
.
tic
()
except
:
raise
StandardError
(
"Could not find ingestion with reference %s"
%
reference
)
def
simulateIngestionAlarm
(
self
,
reference
,
now
):
self
.
portal
.
ERP5Site_stopIngestionList
()
self
.
tic
()
self
.
portal
.
ERP5Site_createDataAnalysisList
()
time
.
sleep
(
5
)
self
.
tic
()
self
.
portal
.
ERP5Site_executeDataAnalysisList
()
time
.
sleep
(
5
)
self
.
tic
()
def
ingest
(
self
,
data_chunk
,
reference
,
extension
):
ingestion_policy
=
self
.
getIngestionPolicy
(
reference
,
self
.
INGESTION_SCRIPT
)
ingestion_reference
=
self
.
getIngestionReference
(
reference
,
extension
)
now
=
datetime
.
now
()
self
.
ingestRequest
(
'POST'
,
(
self
.
USER
,
self
.
PASS
),
ingestion_reference
,
self
.
SINGLE_INGESTION_END
,
data_chunk
,
ingestion_policy
)
ingestion_id
,
ingestion_reference
=
self
.
sanitizeReference
(
ingestion_reference
)
self
.
simulateIngestionAlarm
(
ingestion_reference
,
now
)
return
ingestion_reference
def
checkDataObjects
(
self
,
ingestion_reference
,
data_chunk
,
array
,
json_data
):
self
.
checkOperation
(
None
,
ingestion_reference
,
data_chunk
,
array
,
json_data
)
return
def
checkOperation
(
self
,
ingestion_reference
,
operation_reference
,
data_chunk
,
array
,
json_data
):
if
ingestion_reference
!=
None
:
data_ingestion
=
self
.
getDataIngestion
(
ingestion_reference
)
self
.
assertEqual
(
data_ingestion
,
None
)
data_analysis
=
self
.
getDataAnalysis
(
operation_reference
)
self
.
assertNotEqual
(
data_analysis
,
None
)
data_analysis
=
self
.
getDataAnalysis
(
ingestion_reference
)
self
.
assertEqual
(
data_analysis
,
None
)
data_stream
=
self
.
getDataStream
(
ingestion_reference
)
self
.
assertEqual
(
data_stream
,
None
)
data_array
=
self
.
getDataArray
(
ingestion_reference
)
self
.
assertEqual
(
data_array
,
None
)
data_descriptor
=
self
.
getDataDescriptor
(
ingestion_reference
)
self
.
assertEqual
(
data_descriptor
,
None
)
data_ingestion
=
self
.
getDataIngestion
(
operation_reference
)
self
.
assertEqual
(
data_ingestion
.
getSimulationState
(),
"delivered"
)
size
,
hash_value
=
self
.
context
.
generateSizeHash
(
data_chunk
)
data_stream
=
self
.
getDataStream
(
operation_reference
)
self
.
assertEqual
(
len
(
data_chunk
),
len
(
data_stream
.
getData
()))
self
.
assertEqual
(
size
,
data_stream
.
getSize
())
self
.
assertEqual
(
hash_value
,
data_stream
.
getVersion
())
self
.
assertEqual
(
data_chunk
,
data_stream
.
getData
())
data_array
=
self
.
getDataArray
(
operation_reference
)
if
array
is
None
:
self
.
assertEqual
(
array
,
data_array
.
getArray
())
else
:
np
.
testing
.
assert_allclose
(
array
,
data_array
.
getArray
()[:])
self
.
assertTrue
(
np
.
allclose
(
array
,
data_array
.
getArray
()[:]))
if
ingestion_reference
==
None
:
data_descriptor
=
self
.
getDataDescriptor
(
operation_reference
)
self
.
assertEqual
(
json_data
,
data_descriptor
.
getTextContent
())
def
perform_csv_test
(
self
,
extension
,
delimiter
):
file_name
=
"file_name.csv"
reference
=
self
.
getRandomReference
()
array
=
[[
random
.
random
()
for
i
in
range
(
self
.
CHUNK_SIZE_CSV
+
10
)]
for
j
in
range
(
self
.
CHUNK_SIZE_CSV
+
10
)]
np
.
savetxt
(
file_name
,
array
,
delimiter
=
delimiter
)
chunk
=
[]
with
open
(
file_name
,
'r'
)
as
csv_file
:
data_chunk
=
csv_file
.
read
()
csv_file
.
seek
(
0
)
reader
=
csv
.
reader
(
csv_file
,
delimiter
=
delimiter
)
for
index
,
line
in
enumerate
(
reader
):
if
(
index
<
self
.
CHUNK_SIZE_CSV
):
chunk
.
append
(
line
)
else
:
break
data
=
{}
data
[
"csv"
]
=
chunk
json_data
=
json
.
dumps
(
data
)
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
extension
)
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
array
[:
self
.
CHUNK_SIZE_CSV
],
json_data
)
if
os
.
path
.
exists
(
file_name
):
os
.
remove
(
file_name
)
def
perform_nii_test
(
self
,
gz
):
import
nibabel
as
nib
import
numpy
as
np
reference
=
self
.
getRandomReference
()
url
=
'data_stream_module/mne_sample_for_test_nii'
if
gz
:
url
+=
"_gz"
sample_data_stream
=
self
.
context
.
restrictedTraverse
(
url
)
data_chunk
=
sample_data_stream
.
getData
()
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
GZ
if
gz
else
self
.
NII
)
file_name
=
"file_name.nii"
if
gz
:
file_name
+=
".gz"
with
open
(
file_name
,
'w'
)
as
nii_file
:
nii_file
.
write
(
data_chunk
)
img
=
nib
.
load
(
file_name
)
data
=
{}
for
key
in
img
.
header
:
try
:
if
type
(
img
.
header
[
key
])
is
np
.
ndarray
:
content
=
img
.
header
[
key
].
tolist
()
try
:
if
np
.
isnan
(
img
.
header
[
key
]):
content
=
"nan"
except
:
pass
else
:
content
=
img
.
header
[
key
]
if
content
==
0
:
content
=
"0"
json
.
dumps
(
content
)
if
content
!=
""
:
data
[
key
]
=
content
except
Exception
as
e
:
pass
# ignore non serializable info
json_data
=
json
.
dumps
(
data
)
if
os
.
path
.
exists
(
file_name
):
os
.
remove
(
file_name
)
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
def
test_full_data_ingestion
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
,
nparray
,
json_data
=
self
.
generateRawDataBytesAndArray
()
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
FIF
)
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
nparray
,
json_data
)
def
test_text_data_ingestion
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
TXT
)
json_data
=
json
.
dumps
({
"File content sample: "
:
data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
def
test_tsv_data_ingestion
(
self
):
delimiter
=
"
\
t
"
extension
=
self
.
TSV
self
.
perform_csv_test
(
extension
,
delimiter
)
def
test_csv_data_ingestion
(
self
):
delimiter
=
","
extension
=
self
.
CSV
self
.
perform_csv_test
(
extension
,
delimiter
)
def
test_default_text_data_ingestion
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
RANDOM
)
json_data
=
json
.
dumps
({
"File content sample: "
:
data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
def
test_nii_data_ingestion
(
self
):
self
.
perform_nii_test
(
gz
=
False
)
def
test_nii_gz_data_ingestion
(
self
):
self
.
perform_nii_test
(
gz
=
True
)
def
test_data_ingestion_split_file
(
self
):
reference
=
self
.
getRandomReference
()
ingestion_policy
=
self
.
getIngestionPolicy
(
reference
,
self
.
INGESTION_SCRIPT
)
data_chunk_1
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
250
)])
data_chunk_2
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
250
)])
data_chunk_3
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
250
)])
data_chunk_4
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
250
)])
data_chunk
=
data_chunk_1
+
data_chunk_2
+
data_chunk_3
+
data_chunk_4
ingestion_reference
=
self
.
getIngestionReference
(
reference
,
self
.
FIF
)
self
.
ingestRequest
(
'POST'
,
(
self
.
USER
,
self
.
PASS
),
ingestion_reference
,
self
.
PART_1
,
data_chunk_1
,
ingestion_policy
)
time
.
sleep
(
1
)
self
.
ingestRequest
(
'POST'
,
(
self
.
USER
,
self
.
PASS
),
ingestion_reference
,
self
.
PART_2
,
data_chunk_2
,
ingestion_policy
)
time
.
sleep
(
1
)
self
.
ingestRequest
(
'POST'
,
(
self
.
USER
,
self
.
PASS
),
ingestion_reference
,
self
.
PART_3
,
data_chunk_3
,
ingestion_policy
)
time
.
sleep
(
1
)
self
.
ingestRequest
(
'POST'
,
(
self
.
USER
,
self
.
PASS
),
ingestion_reference
,
self
.
EOF
,
data_chunk_4
,
ingestion_policy
)
self
.
portal
.
ERP5Site_stopIngestionList
()
self
.
tic
()
ingestion_id
,
ingestion_reference
=
self
.
sanitizeReference
(
ingestion_reference
)
data_stream
=
self
.
getDataStream
(
ingestion_reference
)
self
.
assertEqual
(
len
(
data_chunk
),
len
(
data_stream
.
getData
()))
self
.
assertEqual
(
data_chunk
,
data_stream
.
getData
())
def
test_deletion
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
TXT
)
json_data
=
json
.
dumps
({
"File content sample: "
:
data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
self
.
portal
.
ERP5Site_invalidateIngestionObjects
(
ingestion_reference
)
self
.
tic
()
invalid_reference
=
ingestion_reference
+
self
.
INVALID
self
.
checkOperation
(
ingestion_reference
,
invalid_reference
,
data_chunk
,
None
,
json_data
)
def
test_rename
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
TXT
)
json_data
=
json
.
dumps
({
"File content sample: "
:
data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
new_ingestion_reference
=
ingestion_reference
+
self
.
NEW
self
.
portal
.
ERP5Site_renameIngestion
(
ingestion_reference
,
new_ingestion_reference
)
self
.
tic
()
self
.
checkOperation
(
ingestion_reference
,
new_ingestion_reference
,
data_chunk
,
None
,
json_data
)
def
test_reingestion
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
TXT
)
json_data
=
json
.
dumps
({
"File content sample: "
:
data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
new_data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
new_json_data
=
json
.
dumps
({
"File content sample: "
:
new_data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
ingestion_reference
=
self
.
ingest
(
new_data_chunk
,
reference
,
self
.
TXT
)
self
.
checkDataObjects
(
ingestion_reference
,
new_data_chunk
,
None
,
new_json_data
)
def
test_reference_exists
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
TXT
)
json_data
=
json
.
dumps
({
"File content sample: "
:
data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
size
,
hash_value
=
self
.
context
.
generateSizeHash
(
data_chunk
)
full_reference
=
self
.
getFullReference
(
ingestion_reference
,
size
,
hash_value
)
exists
=
self
.
portal
.
ingestionReferenceExists
(
full_reference
)
self
.
assertEqual
(
exists
,
self
.
TRUE
)
def
test_descriptor_html_content_script
(
self
):
reference
=
self
.
getRandomReference
()
data_chunk
=
''
.
join
([
random
.
choice
(
string
.
ascii_letters
+
string
.
digits
)
for
_
in
xrange
(
self
.
CHUNK_SIZE_TXT
+
1000
)])
ingestion_reference
=
self
.
ingest
(
data_chunk
,
reference
,
self
.
TXT
)
json_data
=
json
.
dumps
({
"File content sample: "
:
data_chunk
[:
self
.
CHUNK_SIZE_TXT
]})
self
.
checkDataObjects
(
ingestion_reference
,
data_chunk
,
None
,
json_data
)
script_content
=
self
.
portal
.
getDescriptorHTMLContent
(
ingestion_reference
)
self
.
assertEqual
(
script_content
,
json_data
)
'''
def test_data_stream_invalidation(self):
reference = self.getRandomReference()
data_chunk = ''.join([random.choice(string.ascii_letters + string.digits) for _ in xrange(self.CHUNK_SIZE_TXT + 1000)])
ingestion_reference = self.ingest(data_chunk, reference, self.TXT)
json_data = json.dumps({"File content sample: ": data_chunk[:self.CHUNK_SIZE_TXT]})
self.checkDataObjects(ingestion_reference, data_chunk, None, json_data)
#self.portal.data_stream_module.invalidateObject(data_stream.getId())
self.tic()
invalid_reference = ingestion_reference + self.INVALID
self.checkOperation(ingestion_reference, invalid_reference, data_chunk, None, json_data)
'''
# TODOs
#def test_object_invalidation(self):
#def test_deletion_rename_and_reingestion_on_split_ingestion(self):
#def test_descriptor_html_content_script_on_middle_of_ingestion(self):
#def test_usual_features_on_middle_of_ingestion(self):
bt5/erp5_wendelin_data_lake_ingestion/TestTemplateItem/portal_components/test.erp5.testDataIngestion.xml
deleted
100644 → 0
View file @
7f5cf88b
<?xml version="1.0"?>
<ZopeData>
<record
id=
"1"
aka=
"AAAAAAAAAAE="
>
<pickle>
<global
name=
"Test Component"
module=
"erp5.portal_type"
/>
</pickle>
<pickle>
<dictionary>
<item>
<key>
<string>
_recorded_property_dict
</string>
</key>
<value>
<persistent>
<string
encoding=
"base64"
>
AAAAAAAAAAI=
</string>
</persistent>
</value>
</item>
<item>
<key>
<string>
default_reference
</string>
</key>
<value>
<string>
testDataIngestion
</string>
</value>
</item>
<item>
<key>
<string>
description
</string>
</key>
<value>
<none/>
</value>
</item>
<item>
<key>
<string>
id
</string>
</key>
<value>
<string>
test.erp5.testDataIngestion
</string>
</value>
</item>
<item>
<key>
<string>
portal_type
</string>
</key>
<value>
<string>
Test Component
</string>
</value>
</item>
<item>
<key>
<string>
sid
</string>
</key>
<value>
<none/>
</value>
</item>
<item>
<key>
<string>
text_content_error_message
</string>
</key>
<value>
<tuple/>
</value>
</item>
<item>
<key>
<string>
text_content_warning_message
</string>
</key>
<value>
<tuple>
<string>
W:181, 4: Unused variable \'ingestion_id\' (unused-variable)
</string>
<string>
W:234, 34: Unused variable \'i\' (unused-variable)
</string>
<string>
W:234, 76: Unused variable \'j\' (unused-variable)
</string>
<string>
W:257, 4: Redefining name \'np\' from outer scope (line 9) (redefined-outer-name)
</string>
<string>
W:257, 4: Reimport \'numpy\' (imported line 9) (reimported)
</string>
<string>
W:273, 11: Using type() instead of isinstance() for a typecheck. (unidiomatic-typecheck)
</string>
<string>
W:277, 10: No exception type(s) specified (bare-except)
</string>
<string>
W:285, 26: Unused variable \'e\' (unused-variable)
</string>
<string>
W:350, 4: Unused variable \'ingestion_id\' (unused-variable)
</string>
<string>
W: 8, 0: Unused timedelta imported from datetime (unused-import)
</string>
<string>
W: 10, 0: Unused import math (unused-import)
</string>
<string>
W: 13, 0: Unused log imported from Products.ERP5Type.Log (unused-import)
</string>
<string>
W: 14, 0: Unused import hashlib (unused-import)
</string>
</tuple>
</value>
</item>
<item>
<key>
<string>
version
</string>
</key>
<value>
<string>
erp5
</string>
</value>
</item>
<item>
<key>
<string>
workflow_history
</string>
</key>
<value>
<persistent>
<string
encoding=
"base64"
>
AAAAAAAAAAM=
</string>
</persistent>
</value>
</item>
</dictionary>
</pickle>
</record>
<record
id=
"2"
aka=
"AAAAAAAAAAI="
>
<pickle>
<global
name=
"PersistentMapping"
module=
"Persistence.mapping"
/>
</pickle>
<pickle>
<dictionary>
<item>
<key>
<string>
data
</string>
</key>
<value>
<dictionary/>
</value>
</item>
</dictionary>
</pickle>
</record>
<record
id=
"3"
aka=
"AAAAAAAAAAM="
>
<pickle>
<global
name=
"PersistentMapping"
module=
"Persistence.mapping"
/>
</pickle>
<pickle>
<dictionary>
<item>
<key>
<string>
data
</string>
</key>
<value>
<dictionary>
<item>
<key>
<string>
component_validation_workflow
</string>
</key>
<value>
<persistent>
<string
encoding=
"base64"
>
AAAAAAAAAAQ=
</string>
</persistent>
</value>
</item>
</dictionary>
</value>
</item>
</dictionary>
</pickle>
</record>
<record
id=
"4"
aka=
"AAAAAAAAAAQ="
>
<pickle>
<global
name=
"WorkflowHistoryList"
module=
"Products.ERP5Type.patches.WorkflowTool"
/>
</pickle>
<pickle>
<tuple>
<none/>
<list>
<dictionary>
<item>
<key>
<string>
action
</string>
</key>
<value>
<string>
validate
</string>
</value>
</item>
<item>
<key>
<string>
validation_state
</string>
</key>
<value>
<string>
validated
</string>
</value>
</item>
</dictionary>
</list>
</tuple>
</pickle>
</record>
</ZopeData>
bt5/erp5_wendelin_data_lake_ingestion/bt/template_test_id_list
deleted
100644 → 0
View file @
7f5cf88b
test.erp5.testDataIngestion
\ No newline at end of file
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment