[Feature Request] Parser error reporting

[the-solipsist]

Other {h}ledger-language grammar add-ons (such as the grammar for Atom) recognizes parser errors, and thus prevents you from making mistakes while adding/editing entries.

It'd be great if this could be made an optional feature with an override command.

[mhansen]

A reasonable feature request. How do the other ones notice errors? Are they continuously shelling out to hledger to parse the file?

…

On Mon, 5 Sep 2022 at 19:42, Pranesh Prakash ***@***.***> wrote: Other {h}ledger-language grammar add-ons (such as the grammar for Atom <https://github.com/4ourbit/language-ledger#parser-error-reports>) recognizes parser errors, and thus prevents you from making mistakes while adding/editing entries. It'd be great if this could be made an optional feature with an override command. — Reply to this email directly, view it on GitHub <#231>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAZYOJPC4MMI7UF5LSZ26LV4W575ANCNFSM6AAAAAAQE2DKRI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[the-solipsist]

I'm not a coder, but I believe this is the code which the Atom add-on uses:
https://github.com/4ourbit/language-ledger/blob/master/lib/language-ledger.coffee#L46-L91

[mhansen]

Right, looks like they call “ledger stats” whenever the file changes. My hledger file is a few years of data and takes a few seconds to output. If i set this up to call hledger on every save then my cpu usage would go high and my laptop fan would turn on. Maybe if there was a way to invoke hledger just for the current file looking for syntax errors, rather than balance assertions etc too, could use that.

…

On Mon, 5 Sep 2022 at 21:08, Pranesh Prakash ***@***.***> wrote: I'm not a coder, but I believe this is the code which the Atom add-on uses: https://github.com/4ourbit/language-ledger/blob/master/lib/language-ledger.coffee#L46-L91 — Reply to this email directly, view it on GitHub <#231 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAZYOPABKOL2A3V6KGIYKLV4XIBVANCNFSM6AAAAAAQE2DKRI> . You are receiving this because you commented.Message ID: ***@***.***>

[simonmichael]

Related, see hledger-flycheck. Several seconds would be a problem though. For me it's instant (recent macbook, journal split by year)

[the-solipsist]

And as per @simonmichael, hledger check (or perhaps hledger check parseable) would be the fastest / most appropriate command for checking validity.

[mhansen]

Looks like hledger-flycheck emacs plugin invokes "hledger check" on the shell: https://github.com/DamienCassou/flycheck-hledger/blob/master/flycheck-hledger.el#L63

Thanks for the pointers. Let's try some benchmarks on my shiny new M1 (which should be a bit faster than my other laptop):

$ hyperfine "hledger check" "hledger check parseable" "hledger bal" "hledger stats"
Benchmark #1: hledger check
  Time (mean ± σ):     471.3 ms ±   5.1 ms    [User: 438.8 ms, System: 26.9 ms]
  Range (min … max):   465.7 ms … 477.8 ms    10 runs
 
Benchmark #2: hledger check parseable
  Time (mean ± σ):     471.5 ms ±   6.0 ms    [User: 439.8 ms, System: 26.4 ms]
  Range (min … max):   465.1 ms … 480.2 ms    10 runs
 
Benchmark #3: hledger bal
  Time (mean ± σ):     582.1 ms ±   6.0 ms    [User: 545.0 ms, System: 30.8 ms]
  Range (min … max):   575.4 ms … 593.4 ms    10 runs
 
Benchmark #4: hledger stats
  Time (mean ± σ):     559.5 ms ±   4.8 ms    [User: 527.9 ms, System: 26.1 ms]
  Range (min … max):   555.9 ms … 568.2 ms    10 runs
 
Summary
  'hledger check' ran
    1.00 ± 0.02 times faster than 'hledger check parseable'
    1.19 ± 0.02 times faster than 'hledger stats'
    1.24 ± 0.02 times faster than 'hledger bal'

Looks like hledger check and hledger check parseable are about the same speed, both 19% faster than stats and 24% faster than bal.

Some stats about my journal:

$ hledger stats
...
Transactions span        : 2008-05-19 to 2022-09-04 (5221 days)
Last transaction         : 2022-09-03 (3 days ago)
Transactions             : 15529 (3.0 per day)
...
Run time (throughput)    : 0.57s (27471 txns/s)

I imagine it'd be near-instant if I split by year. I don't really want to split by year though.

Maybe to be quick, an extension could use hledger check on the currently-edited individual file along with --ignore-assertions, because none of my individual files balance correctly without other files. I was a bit surprised that check and check parseable apply the balance assertions (I would have guessed that's beyond the scope of parsing?).

[simonmichael]

The (3?) "Basic assertions" are always run, currently it's not possible to separate them.