ylqjgm
/
canvas-lms
mirror of https://github.com/instructure/canvas-lms.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Have grade calc use INSERT ... ON CONFLICT 

Have the grade calculator update the scores and score_metadata tables
using a single INSERT ... ON CONFLICT DO UPDATE statement rather than
consecutive UPDATE and INSERT statements. This will (we hope) prevent
separate invocations of the grade calculator from stepping on each
other's toes and interrupting calculations midway through due to
constraint errors.

fixes GRADE-2163

Test plan:
- Specs pass
- Set up a course with some assignment groups and grading periods
  - Assign some grades
  - Check that scores are being calculated properly and no errors

Change-Id: I894d5040f8a536c4b91d344dffd246fb0d9a6311
Reviewed-on: https://gerrit.instructure.com/192330
Tested-by: Jenkins
Reviewed-by: Keith Garner <kgarner@instructure.com>
QA-Review: Derek Bender <djbender@instructure.com>
Product-Review: Keith Garner <kgarner@instructure.com>
This commit is contained in:
Adrian Packel 2019-05-06 17:15:24 -05:00
parent 81f7c56011
commit 4aae92655d
1 changed files with 63 additions and 95 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

158
lib/grade_calculator.rb
View File

@ -457,9 +457,9 @@ class GradeCalculator
def update_values_for(column, updates: {}, key: :grade)
return unless column
actual_updates = user_specific_updates(updates: updates, default_value: column, key: key)
actual_updates = user_specific_updates(updates: updates, default_value: "excluded.#{column}", key: key)
"#{column} = CASE enrollment_id #{actual_updates} END"
"#{column} = CASE excluded.enrollment_id #{actual_updates} END"
end
def insert_values_for(column, updates: {}, key: :grade)
@ -500,19 +500,17 @@ class GradeCalculator
def save_course_and_grading_period_scores
return if @only_update_course_gp_metadata
# Construct upsert statement to update existing Scores or create them if needed,
# for course and grading period Scores.
# Depending on whether we're updating course scores or grading period
# scores, we need to check our inserted values against different uniqueness
# constraints
conflict_target = if @grading_period.present?
"(enrollment_id, grading_period_id) WHERE grading_period_id IS NOT NULL"
else
"(enrollment_id) WHERE course_score"
end
# Update existing course and grading period Scores or create them if needed.
Score.connection.execute("
UPDATE #{Score.quoted_table_name}
SET
#{columns_to_insert_or_update[:update_values].join(', ')},
updated_at = #{updated_at},
-- if workflow_state was previously deleted for some reason, update it to active
workflow_state = COALESCE(NULLIF(workflow_state, 'deleted'), 'active')
WHERE
enrollment_id IN (#{joined_enrollment_ids}) AND
assignment_group_id IS NULL AND
#{@grading_period ? "grading_period_id = #{@grading_period.id}" : 'course_score IS TRUE'};
INSERT INTO #{Score.quoted_table_name}
(
enrollment_id, grading_period_id,
@ -527,13 +525,14 @@ class GradeCalculator
#{updated_at} as created_at,
#{updated_at} as updated_at
FROM #{Enrollment.quoted_table_name} enrollments
LEFT OUTER JOIN #{Score.quoted_table_name} scores on
scores.enrollment_id = enrollments.id AND
scores.assignment_group_id IS NULL AND
#{@grading_period ? "scores.grading_period_id = #{@grading_period.id}" : 'scores.course_score IS TRUE'}
WHERE
enrollments.id IN (#{joined_enrollment_ids}) AND
scores.id IS NULL;
enrollments.id IN (#{joined_enrollment_ids})
ON CONFLICT #{conflict_target}
DO UPDATE SET
#{columns_to_insert_or_update[:update_values].join(', ')},
updated_at = excluded.updated_at,
-- if workflow_state was previously deleted for some reason, update it to active
workflow_state = COALESCE(NULLIF(excluded.workflow_state, 'deleted'), 'active')
")
end
@ -545,23 +544,6 @@ class GradeCalculator
return unless @ignore_muted
ScoreMetadata.connection.execute("
UPDATE #{ScoreMetadata.quoted_table_name} metadata
SET
calculation_details = CASE enrollments.user_id
#{@dropped_updates.map do |user_id, dropped|
"WHEN #{user_id} THEN cast('#{dropped.to_json}' as json)"
end.join(' ')}
ELSE calculation_details
END,
updated_at = #{updated_at}
FROM #{Score.quoted_table_name} scores
INNER JOIN #{Enrollment.quoted_table_name} enrollments ON
enrollments.id = scores.enrollment_id
WHERE
scores.enrollment_id IN (#{joined_enrollment_ids}) AND
scores.assignment_group_id IS NULL AND
#{@grading_period ? "scores.grading_period_id = #{@grading_period.id}" : 'scores.course_score IS TRUE'} AND
metadata.score_id = scores.id;
INSERT INTO #{ScoreMetadata.quoted_table_name}
(score_id, calculation_details, created_at, updated_at)
SELECT
@ -582,77 +564,73 @@ class GradeCalculator
WHERE
scores.enrollment_id IN (#{joined_enrollment_ids}) AND
scores.assignment_group_id IS NULL AND
#{@grading_period ? "scores.grading_period_id = #{@grading_period.id}" : 'scores.course_score IS TRUE'} AND
metadata.id IS NULL;
#{@grading_period ? "scores.grading_period_id = #{@grading_period.id}" : 'scores.course_score IS TRUE'}
ON CONFLICT (score_id)
DO UPDATE SET
calculation_details = excluded.calculation_details,
updated_at = excluded.updated_at
;
")
end
def assignment_group_columns_to_insert_or_update
return @assignment_group_columns_to_insert_or_update if defined? @assignment_group_columns_to_insert_or_update
column_list = { value_names: [], update_columns: [], update_values: [], insert_columns: [], insert_values: [] }
column_list = {
insert_columns: [],
insert_values: [],
update_columns: [],
update_values: [],
value_names: []
}
unless @only_update_points
column_list[:value_names] << 'current_score'
column_list[:update_columns] << "#{current_score_column} = val.current_score"
column_list[:update_columns] << "#{current_score_column} = excluded.current_score"
column_list[:insert_columns] << "val.current_score AS #{current_score_column}"
column_list[:value_names] << 'final_score'
column_list[:update_columns] << "#{final_score_column} = val.final_score"
column_list[:update_columns] << "#{final_score_column} = excluded.final_score"
column_list[:insert_columns] << "val.final_score AS #{final_score_column}"
end
column_list[:value_names] << 'current_points'
column_list[:update_columns] << "#{points_column(:current)} = val.current_points"
column_list[:update_columns] << "#{points_column(:current)} = excluded.current_points"
column_list[:insert_columns] << "val.current_points AS #{points_column(:current)}"
column_list[:value_names] << 'final_points'
column_list[:update_columns] << "#{points_column(:final)} = val.final_points"
column_list[:update_columns] << "#{points_column(:final)} = excluded.final_points"
column_list[:insert_columns] << "val.final_points AS #{points_column(:final)}"
@assignment_group_columns_to_insert_or_update = column_list
end
def save_assignment_group_scores(score_values, dropped_values)
# Construct upsert statement to update existing Scores or create them if needed,
# for assignment group Scores.
# Update existing assignment group Scores or create them if needed.
Score.connection.execute("
UPDATE #{Score.quoted_table_name} scores
SET
#{assignment_group_columns_to_insert_or_update[:update_columns].join(', ')},
updated_at = #{updated_at},
workflow_state = COALESCE(NULLIF(workflow_state, 'deleted'), 'active')
FROM (VALUES #{score_values}) val
(
enrollment_id,
assignment_group_id,
#{assignment_group_columns_to_insert_or_update[:value_names].join(', ')}
)
WHERE val.enrollment_id = scores.enrollment_id AND
val.assignment_group_id = scores.assignment_group_id;
INSERT INTO #{Score.quoted_table_name} (
enrollment_id, assignment_group_id,
#{assignment_group_columns_to_insert_or_update[:value_names].join(', ')},
course_score, created_at, updated_at
enrollment_id, assignment_group_id,
#{assignment_group_columns_to_insert_or_update[:value_names].join(', ')},
course_score, created_at, updated_at
)
SELECT
val.enrollment_id AS enrollment_id,
val.assignment_group_id as assignment_group_id,
#{assignment_group_columns_to_insert_or_update[:insert_columns].join(', ')},
FALSE AS course_score,
#{updated_at} AS created_at,
#{updated_at} AS updated_at
FROM (VALUES #{score_values}) val
(
enrollment_id,
assignment_group_id,
#{assignment_group_columns_to_insert_or_update[:value_names].join(', ')}
)
SELECT
val.enrollment_id AS enrollment_id,
val.assignment_group_id as assignment_group_id,
#{assignment_group_columns_to_insert_or_update[:insert_columns].join(', ')},
FALSE AS course_score,
#{updated_at} AS created_at,
#{updated_at} AS updated_at
FROM (VALUES #{score_values}) val
(
enrollment_id,
assignment_group_id,
#{assignment_group_columns_to_insert_or_update[:value_names].join(', ')}
)
LEFT OUTER JOIN #{Score.quoted_table_name} sc ON
sc.enrollment_id = val.enrollment_id AND
sc.assignment_group_id = val.assignment_group_id
WHERE
sc.id IS NULL;
ON CONFLICT (enrollment_id, assignment_group_id) WHERE assignment_group_id IS NOT NULL
DO UPDATE SET
#{assignment_group_columns_to_insert_or_update[:update_columns].join(', ')},
updated_at = excluded.updated_at,
workflow_state = COALESCE(NULLIF(excluded.workflow_state, 'deleted'), 'active')
")
# We only save score metadata for posted grades. This means, if we're
@ -661,17 +639,6 @@ class GradeCalculator
# score metadata for unposted grades.
if @ignore_muted
ScoreMetadata.connection.execute("
UPDATE #{ScoreMetadata.quoted_table_name} md
SET
calculation_details = CAST(val.calculation_details as json),
updated_at = #{updated_at}
FROM (VALUES #{dropped_values}) val
(enrollment_id, assignment_group_id, calculation_details)
INNER JOIN #{Score.quoted_table_name} scores ON
scores.enrollment_id = val.enrollment_id AND
scores.assignment_group_id = val.assignment_group_id
WHERE
scores.id = md.score_id;
INSERT INTO #{ScoreMetadata.quoted_table_name}
(score_id, calculation_details, created_at, updated_at)
SELECT
@ -684,10 +651,11 @@ class GradeCalculator
LEFT OUTER JOIN #{Score.quoted_table_name} scores ON
scores.enrollment_id = val.enrollment_id AND
scores.assignment_group_id = val.assignment_group_id
LEFT OUTER JOIN #{ScoreMetadata.quoted_table_name} metadata ON
metadata.score_id = scores.id
WHERE
metadata.id IS NULL;
ON CONFLICT (score_id)
DO UPDATE SET
calculation_details = excluded.calculation_details,
updated_at = excluded.updated_at
;
")
end
end