Re: [sqlite] How does one block a reader connection?

2017-03-26 Thread Rowan Worth
On 26 March 2017 at 14:17, Keith Medcalf wrote: > If you do not specify your own custom busy handler (to display flying ball > bearings, etc, or do your own exponential sleeping, etc) then the default > busy_handler is used. The default busy handler does its own exponential

Re: [sqlite] table-naming-expression impact on sqlite3_prepare

2017-03-26 Thread Simon Slavin
On 27 Mar 2017, at 1:52am, Kees Nuyt wrote: > It's something that can be done by any host language. No need to implement > that in SQL. Also, you’re scripting a shell tool. So write a text file with your SQL commands in and feed it to the shell tool whole. Simon.

Re: [sqlite] table-naming-expression impact on sqlite3_prepare

2017-03-26 Thread Kees Nuyt
On Sun, 26 Mar 2017 15:34:22 -0700, petern wrote: > Here is your suggestion with matched brackets and quotes and assuming > mytable has a column [tablename]: > > select eval(printf('create table %s (a,b,c)',tablename)) from mytables; I think you mean:

Re: [sqlite] table-naming-expression impact on sqlite3_prepare

2017-03-26 Thread E.Pasma
27 mrt 2017, petern: In general I've been thinking about materializing data dependent temporary tables and even using them in CTE's. The tremendous expressive economy of TCL and somewhat built-in support within SQLite got me thinking. Consider the problem of pivot table function for

Re: [sqlite] Syntax. table-function-name vs table-naming-function-name

2017-03-26 Thread petern
Richard, thank you for your reply. I really appreciate it. The fact that you have carefully thought about how to cross the FROM clause barrier with expressions is itself a useful fact. If you say the current implementation is painted into a corner on this issue I believe you. It would be

Re: [sqlite] table-naming-expression impact on sqlite3_prepare

2017-03-26 Thread petern
That you for your kind comment about my table-naming-expression proposal. Here is your suggestion with matched brackets and quotes and assuming mytable has a column [tablename]: select eval(printf('create table %s (a,b,c)',tablename)) from mytables; Significant credit should also go to the

Re: [sqlite] Syntax. table-function-name vs table-naming-function-name

2017-03-26 Thread Richard Hipp
On 3/25/17, petern wrote: > > Why can't we have a parallel syntax branch for scalar valued > "table-naming-function-name"? In other words, why not have support for > simply naming an existing table or view by return value of a scalar > function? > The easiest way

Re: [sqlite] table-naming-expression impact on sqlite3_prepare

2017-03-26 Thread E.Pasma
26-03-2017 petern : > The table-naming-expression, if > normal expressions are allowed, would obviously require sqlite3_prepare to > consult the database in situations where the name string expression depended > on a SQL statement being evaluated. Is this the main problem with allowing >

[sqlite] table-naming-expression impact on sqlite3_prepare

2017-03-26 Thread petern
My thanks to everyone who responded to my read blocking transaction isolation question. Further to my other question/proposal with no responses, what would be the impact on sqlite3_prepare to introduce a new branch called table-naming-expression into the syntax graph at:

Re: [sqlite] How does one block a reader connection?

2017-03-26 Thread petern
Keith, I understand your point. The timescale of polling is between 1 and 10 seconds by sleep loop depending on operational objectives. This range of sleep loop will have a corresponding latency of between 0.5 and 5 seconds for single commands with a uniform arrival time distribution. The idea

Re: [sqlite] 3.17.0 bug report: FTS5 insertion puts a wrong value in last_insert_rowid

2017-03-26 Thread Florian Weimer
* Gwendal Roué: > I have found a regression in SQLite 3.17.0. In the following SQL statements: > > CREATE VIRTUAL TABLE t1 USING FTS5(content); > INSERT INTO t1(content) VALUES ('some text'); > SELECT last_insert_rowid(); // 10 (wrong) > SELECT rowid FROM t1; // 1 > > The

[sqlite] 3.17.0 bug report: FTS5 insertion puts a wrong value in last_insert_rowid

2017-03-26 Thread Gwendal Roué
Hello, I have found a regression in SQLite 3.17.0. In the following SQL statements: CREATE VIRTUAL TABLE t1 USING FTS5(content); INSERT INTO t1(content) VALUES ('some text'); SELECT last_insert_rowid(); // 10 (wrong) SELECT rowid FROM t1; // 1 The expected value of the the

Re: [sqlite] How does one block a reader connection?

2017-03-26 Thread Jens Alfke
> On Mar 25, 2017, at 3:52 PM, petern wrote: > > So finally, here is the question. Is there a SQLite API way for reader > connections to block and wait for a meaningful change, like a new row, in > the 'cmd' table instead of madly polling and using up database

Re: [sqlite] How does one block a reader connection?

2017-03-26 Thread Keith Medcalf
Saturday, 25 March, 2017 23:44. petern wrote: > Can anybody explain the purpose of > http://sqlite.org/c3ref/busy_handler.html > ? It seems the only practical use would be to allow the caller to give > the engine a suggested lock deadline before SQLITE_BUSY is