Command line improvements

[MichaelDBA]

First of all, thanks for your new tool! Here are a few suggestions:

1. Make table output optional or filterable through a new command parameter. Thousands of tables within a database can cause a lot of unwanted output, perhaps something like --no-tables.

2. Add a schema option to limit table output, --schema schema1, schema2

3. Put system information output before table output. That way top level stuff is always at top.

Keep up the good work!

[mdevan]

Thanks, good suggestions!

How about a couple of options like:

--match-table=PATTERN    include only tables whose name matches PATTERN
--match-schema=PATTERN   include objects only from schema that match PATTERN

where PATTERN is a POSIX regular expression, and if it starts with ! then the meaning is inverted (exclude tables/schema). Both options may be used at the same time and works like "AND".

Do you think this will fit your use cases?

[MichaelDBA]

The regex stuff is more than I asked for, but sure that'll work for most table and schema filtering, but what about the case where I don't want table info? Now I have to write a regex pattern that will not return anything as opposed to a parameter that just says --no-tables and/or --no-schemas. How about both? Also, if I use regex filtering it makes it more difficult than to specify 2 schemas directly with the --schema parameter which expects a comma-delimited list of schemas.

It seems to me regex stuff is more applicable to tables where you could have thousands, but not to schemas where you might have a few. So I would recommend the regex for tables but a comma-delimited list for schemas.

Also, don't forget about the system info being placed before table output.

2 ways to specify schemas:

--schemas schema1, schema2
--schema schema1 --schema schema2

[mdevan]

OK, let's list down the items.

1. filter tables: use an option to include/exclude tables based on regexps
2. filter schema: use a repeatable option to only include objects from the specified schema

These two sound good, let's go with it.

I guess the third item is what objects we should collect from a single database. Right now, pgmetrics gets information about:

• tables
• indexes
• sequences
• user function timings
• extensions (this one is not schema-specific)
• disabled triggers

So if we want to collect only some of these types, then there should be an option that specifies that, I think. Like "--collect=tables,indexes" would collect only tables and indexes, or "--collect=all" would collect everything. Thoughts?

And the fourth one is about the report cosmetics, which we'll handle separately in another issue -- it is not related to data collection after all. (And I agree, it makes sense to put it just below the "PostgreSQL Cluster:" section. It gives more context for the rest of the report.)

[mdevan]

b38c9ec includes these schema and table selection options:

Collection options:
  -c, --schema=REGEXP          collect only from schema(s) matching POSIX regexp
  -C, --exclude-schema=REGEXP  do NOT collect from schema(s) matching POSIX regexp
  -a, --table=REGEXP           collect only from table(s) matching POSIX regexp
  -A, --exclude-table=REGEXP   do NOT collect from table(s) matching POSIX regexp

These are similar to pg_dump's options, except that it's based on regexp.

[mdevan]

13393c2 shows the system information earlier on in the output.

[mdevan]

51462e3 includes option to omit the collection of one or more of: tables, indexes, sequences, functions, extensions and triggers.

      --omit=WHAT              do NOT collect the items specified as a comma-separated
                                   list of: "tables", "indexes", "sequences",
                                   "functions", "extensions", "triggers"

That addresses all the points in this issue, I think. Feel free to reopen this/open another issue if it doesn't.