_SITE_POOLS[frappe.local.site].read_only will hold ConnectionPool for
replica database connections. _SITE_POOLS[frappe.local.site].default
will hold pool for connections that allow read+writes
Start initial pool of 4 (_POOL_SIZE) connections for a given site. When
the pool is exhausted, generate new connections upon request and add
them to the pool. Max allowed size for each pool is 64 (_MAX_POOL_SIZE)
connections. However, you may have more than 64 active connections, just
that they won't be pooled but destroyed upon job completion/end of
request/similar cycle.
* Make all methods static
* Add typing hints
* Don't safe fetch attribute value for errno - MariaDB exceptions if
raised will have errno attr in them. If the class doesn't have one,
it's not meant to be passed in these methods.
Move to use MariaDB's official Python client written in C instead of
the PyMySQL library. This change doesn't rid Frappe of the PyMySQL
library. Instead, it continues to utilize it for the ER module and
converter methods until the MariaDB library adds support for the same.
Ticket: https://jira.mariadb.org/projects/CONPY/issues/CONPY-203
* Added provision for semi-implemneted version
* Hard to fix it completely given it's broken on develop / desk
* Added TODO for adding nestedset related filters
* fix: misc fixes
local.x gets resetted on every request so switched to a simple dict
simplified is_val_used in set_next_val function for sequences
* chore: use multisql for sequence methods
* fix: fields not updating on form
* minor(base_input): removed unnecessary branching in update_input
* chore: remove prints and rename autoincremented_status_map
* chore: added proper type hint + comment + formatting
* fix: added searching in cast_name rather than handling it manually
* fix: share condition query + test_build_match_conditions
* fix: add cast_name to more places
* test: test for sequence
* fix: sequence functions
* fix: inherit frappetestcase
* minor: attach sequence methods to db context local
* chore: update sequence function names in Database
use frappe.db for sequences in naming.py
* fix: convert filename to str (for autoincremented doctypes)
* chore: better regex for modifying values for postgres
* minor: allow changing name column type (if no data is present in the doctype)
* refactor: validate_autoname
converted it to a simple function
enabled changing autoincrement autoname from customize form
* fix: use sql_ddl for change_column_type in postgres
* fix: use not null constraint in postgres when changing name type
* fix(test): updated test_autoincremented_doctype_transition with transitioning when no data is present
* fix(test): updated test_cast_name
probably messed up during rebase
* fix(test): used rollback upon error in transaction for postgres
* chore: use frappe.db.x methods for sequences
* minor: use temporary sequences in test
* minor: use generate_hash for sequence naming in sequence tests
* chore: replace sequence imports with frappe.db.x
* chore: move out casting name fields to a separate method
* refactor: cast_name
more explicit cases for casts and added docstring
* fix: added space in test_cast_name
* chore: fix linter
* chore: better naming for can_change_name_column_type
* chore: add comment for autoincremented_site_status_map
* chore: update/add docstrings
## Issue
`frappe.db.bulk_insert` is not working as expected:
- It will not insert any row if there are less than 3 values
- It will not add 1st row at all. e.g if I'm adding 5 values, it will only add 4.
- It will add values one by one after 2 values, instead it should have inserted items (in db) in chunk of 10000 (as per the code written before).
## Changes Made
- Solved above issues
- use better way to chunk list
- Added Postgres support for bulk_insert API
And now `bulk_insert` will only do **1 db call for each 10000** values.
Note: For testing purpose I made `Test Bulk Insert` doctype and keep chunk size of 100.
## Before
## After
nodocs
