3 ms·

I liked this idea in principle but in practice I found there's just too many debugPrint() lines cluttering the code, and too much overwhelming console spam, so

by gary_0 5y ago

I liked this idea in principle but in practice I found there's just too many debugPrint() lines cluttering the code, and too much overwhelming console spam, so it's not worth it to keep around debug logging (or taking the time to decide what is worthwhile to keep). When I fix a problem gnarly enough to require lots of logging everywhere, once I've tidied up I write a detailed comment or other documentation, and/or use unit tests to make sure the problem stays solved. The logging code doesn't get committed, which also helps keep the fixed code clear in the diff. For the kinds of programs I usually write, I've settled on only using printing/logging for major events and for things that aren't errors but might require developer attention.

supermatt 5y ago
You can often instrument code at build time to facilitate tracing, keeping your code free from much of the logging statements while still providing logs on function calls, arguments and return values. Obviously this (usually) isn't statement-level logging, but its often enough to diagnose issues.
loganhood 5y ago
> in practice I found there's just too many debugPrint() lines cluttering the code On this particular aspect - I find that log lines can do double-duty as code comments. When I see some comment that explains the what-and-why of a section of code - "// We also need to check the Foo state for consistency before committing Bar" - I might actually change it to a log line like - "Verifying state of Foo ${foo.id} before committing Bar ${bar.id}".
tialaramex 5y ago
> overwhelming console spam The article says to use logging and not actual printf debugging, which solves this. If your logging framework is even somewhat capable and modern it understands all of: - named priorities: I want that FATAL log highlighted, but I probably don't want the DEBUG log shown at all unless I am currently debugging this - automatic categories: os.mutex.internal and account.forex.hack can be automatically distinguished in most modern languages by the fact they live in different objects / source code files / even in C we know what the name of the current function is and can at least annotate that in logs. - how to send the log messages somewhere useful, hopefully at least files and some primitive syslog type protocol, maybe a lot more In a nice shiny modern logging framework I'd expect to be provided with default annotations so that e.g. when I error.log("Can't enrol students on a module that doesn't exist") the framework annotates the log entry with: The HTTP request that got us here, including e.g. query parameters and date-time stamp, and the Module we were trying to call a method on when it blew up.
GnarfGnarf 5y ago
Yeah but... aren't you tempted to leave the debugging code, in case you might need it again later?
- gary_0 5y ago
  To me it seems like if it's really that important, it becomes error checking or assertion code, or the code should be refactored to better enforce invariants, or the debug code is spun off into unit tests. That way whatever problem I was trying to detect is checked automatically in the future. And I admit an aesthetic preference for less logging code scattered about. But I haven't found it very useful to barf lots of data into a debug log and then hope to find something useful sifting through it when there's a problem. Maybe it's just one too many information-dense things to juggle for me and smarter people have more luck with this approach. I'm not totally against the idea, I just haven't found a way that works for me.
  - GnarfGnarf 4y ago
    This is where you need to get creative: save the log as HTML, and use colors, tables, etc. to organize the data, catch error needles amongst the data haystack, etc. I find that having lots of trace data is useful because at a glance, you can compare problem situations against normal profiles and patterns.

by gary_0 5y ago

supermatt 5y ago

You can often instrument code at build time to facilitate tracing, keeping your code free from much of the logging statements while still providing logs on function calls, arguments and return values. Obviously this (usually) isn't statement-level logging, but its often enough to diagnose issues.

loganhood 5y ago

> in practice I found there's just too many debugPrint() lines cluttering the code On this particular aspect - I find that log lines can do double-duty as code comments. When I see some comment that explains the what-and-why of a section of code - "// We also need to check the Foo state for consistency before committing Bar" - I might actually change it to a log line like - "Verifying state of Foo ${foo.id} before committing Bar ${bar.id}".

tialaramex 5y ago

> overwhelming console spam The article says to use logging and not actual printf debugging, which solves this. If your logging framework is even somewhat capable and modern it understands all of: - named priorities: I want that FATAL log highlighted, but I probably don't want the DEBUG log shown at all unless I am currently debugging this - automatic categories: os.mutex.internal and account.forex.hack can be automatically distinguished in most modern languages by the fact they live in different objects / source code files / even in C we know what the name of the current function is and can at least annotate that in logs. - how to send the log messages somewhere useful, hopefully at least files and some primitive syslog type protocol, maybe a lot more In a nice shiny modern logging framework I'd expect to be provided with default annotations so that e.g. when I error.log("Can't enrol students on a module that doesn't exist") the framework annotates the log entry with: The HTTP request that got us here, including e.g. query parameters and date-time stamp, and the Module we were trying to call a method on when it blew up.

GnarfGnarf 5y ago

Yeah but... aren't you tempted to leave the debugging code, in case you might need it again later?

gary_0 5y ago
To me it seems like if it's really that important, it becomes error checking or assertion code, or the code should be refactored to better enforce invariants, or the debug code is spun off into unit tests. That way whatever problem I was trying to detect is checked automatically in the future. And I admit an aesthetic preference for less logging code scattered about. But I haven't found it very useful to barf lots of data into a debug log and then hope to find something useful sifting through it when there's a problem. Maybe it's just one too many information-dense things to juggle for me and smarter people have more luck with this approach. I'm not totally against the idea, I just haven't found a way that works for me.
- GnarfGnarf 4y ago
  This is where you need to get creative: save the log as HTML, and use colors, tables, etc. to organize the data, catch error needles amongst the data haystack, etc. I find that having lots of trace data is useful because at a glance, you can compare problem situations against normal profiles and patterns.